Latest / Agentic Conversations (formally mlops.community) / Building a Culture of Experimentation to Speed Up Data-Driven Value // Delina Ivanova // MLOps Coffee Sessions #106

Transcript

Building a Culture of Experimentation to Speed Up Data-Driven Value // Delina Ivanova // MLOps Coffee Sessions #106 · Agentic Conversations (formally mlops.community)

• 0:00
  I like my coffee. Black Lena just blew my mind.
• 0:04
  Who is the Lena? And what are we talking about
• 0:07
  today? So, we had on today, Lena
• 0:10
  Ivanova, who is an associate director for data and insights
• 0:14
  that hellofresh Canada. There.
• 0:16
  She leaves a full-service data team that includes data
• 0:19
  engineering data science and business intelligence, and
• 0:21
  automation, then she's an incredibly personal.
• 0:24
  Incredibly impressive person with a long track record of
• 0:28
  individual contributor work. Work.
• 0:30
  But also technical management work, which we talked a lot
• 0:32
  about Demetrios, isn't that, right?
• 0:34
  So, yeah, I think my favorite thing here was her talking about
• 0:40
  the differences and the advantages that you look at,
• 0:46
  when you're put into a managerial position for a data
• 0:50
  team, the things that she needs to basically block for for her
• 0:55
  team to make sure that her team does what it wants to do.
• 1:00
  So she said and then what she has to think about in return,
• 1:05
  right? Like so she's focusing on how.
• 1:08
  Okay, if this is going to scale later on, what does that look
• 1:11
  like? Also talking to the different
• 1:13
  business stakeholders and explaining what they're
• 1:17
  choosing, why they chose something?
• 1:19
  What the whole rationale behind their choices are and then
• 1:25
  allowing her team to do the fun stuff which is hands on the
• 1:29
  keyboard and Building the actual product or or maintaining it.
• 1:34
  Yeah, I mean I really enjoyed that part of the conversation.
• 1:36
  I'd say I have two really key takeaways that I would encourage
• 1:40
  everybody to watch out for number one.
• 1:41
  I really enjoyed hearing about how complicated hello versus
• 1:45
  business model is, you know, it's easy for us at least in the
• 1:47
  u.s. to just see it as a meal prep company or a meal meal kit
• 1:51
  company but Delena is positioning of it as a food
• 1:55
  solution company. I thought was much more
• 1:56
  interesting and shows how complex the Tire company is and
• 2:00
  how data has to flow through to make them a valuable company,
• 2:03
  which they are. And then second, I thought that
• 2:07
  her her thoughtful Reflections on management or really helpful
• 2:12
  for me to understand, like what goes through the mind of a good
• 2:15
  manager. It helps me as a current
• 2:17
  individual contributor, understand like, what kinds of
• 2:19
  things my manager assistant thinks about and how I can help
• 2:22
  them understand those things. Especially her point about the
• 2:25
  fact that her team will always be more have Expertise in the
• 2:30
  problem that were solving than her.
• 2:32
  I thought was a really humble way to think about management.
• 2:35
  And, you know, I think those kinds of insights are worth
• 2:38
  gold. And so I really enjoyed having
• 2:40
  her on and we don't know this because neither of us have
• 2:44
  worked with her, but I get the feeling that she would be
• 2:46
  awesome to work for. Absolutely.
• 2:49
  I mean, like, yeah, you just I mean talking to her and how she
• 2:53
  speaks about a her team and be what she thinks about and how
• 2:58
  she thinks through His problems and challenges.
• 3:01
  What ever she comes into contact with it's like, oh, wow.
• 3:06
  Like if every manager that I ever had to work for was like
• 3:10
  that. I think it would make work a
• 3:13
  much different field of play. Like it wouldn't be as much of a
• 3:19
  burden as it has been in the past when I've worked for crappy
• 3:23
  managers and I'm sure you've had some pretty bad ones too.
• 3:25
  I think that's just kind of a rite of passage and so when you
• 3:28
  do get a manager, ER, like her. It's, yeah, it's incredible.
• 3:34
  And they really shine. So everybody, if you do work for
• 3:39
  her, please reach out. And let us know if she is as
• 3:42
  awesome as we think she is. And that is it.
• 3:45
  The last thing that I'll say before we jump into the full
• 3:48
  conversation, we're looking for people to help us organize local
• 3:51
  meet up. So if you would like to have a
• 3:54
  local MLA Ops meet up in your city, in real life, we've got
• 3:59
  them. Started in Berlin.
• 4:00
  We've got him started in Amsterdam Barcelona Lisbon,
• 4:04
  there's some in New York San Francisco, we're really looking
• 4:07
  for people in Austin Texas because I hear that's a cool
• 4:11
  place to go and I'm going to use it as an excuse to go there and
• 4:14
  Denver Colorado and Seattle Washington.
• 4:18
  So if you are in any of these places reach out to us and you
• 4:20
  want to organize some stuff and get some cool talks going, let
• 4:23
  us know. Without further Ado, let's get
• 4:26
  into this conversation with Delina.
• 4:33
  First and foremost, I would love to start with career
• 4:37
  trajectories and career growth. I know it's one of your favorite
• 4:42
  topics Delina so we can start start there and maybe it would
• 4:47
  be good just to hear a little bit about your own trajectory
• 4:51
  and how you have progressed through the corporate schema.
• 4:56
  Yeah, for sure. And I think it's a little bit
• 4:58
  different for different people. But one thing that I love about
• 5:02
  data, the data world and everything to do with machine
• 5:05
  learning and data science is that, it's actually quite
• 5:07
  accessible as a field and I love that many people with different
• 5:11
  backgrounds. Can not only learn the skills
• 5:13
  that they need, but they have access to roles that can help
• 5:16
  them get into this field. And in fact, what I often tell
• 5:19
  people is not to Discount any of the previous experience.
• 5:22
  They have because it usually becomes quite valuable when you
• 5:25
  are a data scientist or you're working with machine learning or
• 5:28
  you're even leading a data team, like I am now.
• 5:29
  Now, so for myself, I didn't start out in specifically a data
• 5:34
  science role, but I started working in finance.
• 5:36
  So very again, data heavy, like data, have you roles in a little
• 5:39
  roles but specifically in finance, Financial modeling
• 5:43
  forecasts and did some product analytics and then I actually
• 5:45
  did some Consulting where I looked at operations and
• 5:48
  strategy and I would say that those types of roles really
• 5:52
  taught me a lot about being at solving problems and just
• 5:54
  thinking objectively about problems, learning, how business
• 5:58
  operates, how business Makes decisions and I think these are
• 6:01
  critical skills and knowledge areas that anyone should have to
• 6:05
  Be an Effective data person and be effective at finding Data
• 6:10
  Solutions for business problems. And then I started when they
• 6:13
  design sort of became more of a focus.
• 6:16
  I was working at the bank at the time and we had just stood up
• 6:18
  like an analytics function that guy Central analytics function
• 6:22
  and so I started working there and I started exploring
• 6:25
  initially, some questions, around data ethics.
• 6:27
  So some of the first roll, some Of the first rules that I did
• 6:30
  were really around, data policy data governance.
• 6:33
  But then also data ethics, like how do we use data for good?
• 6:37
  What should we be using versus not using?
• 6:39
  And how do we find kind of creative ways to use data?
• 6:43
  That's not going to create any sort of discriminatory outputs
• 6:47
  or, you know, give it the wrong biasing on the wrong things.
• 6:50
  So to say, which I think sort of set me up to again, be able
• 6:54
  think to think about data problems holistically.
• 6:56
  So both from a problem-solving perspective, but also from A
• 7:00
  perspective of what should we be doing with this?
• 7:02
  Shouldn't we be doing with this? And then that led me to
• 7:05
  eventually to this rule here, where I'm now, hello fresh.
• 7:08
  And now I leave sort of a full data team but I think all of
• 7:11
  these different experiences have made me more effective in this
• 7:14
  role from the perspective of being able to recognize business
• 7:18
  problems. Identifying the right solution.
• 7:21
  We often hear that, you know, prescriptive machine learning
• 7:23
  model isn't always the best solution, sometimes you just
• 7:26
  need, you know, a simple report. So I think those kind of
• 7:29
  Appearances for me combined have helped me be effective here and
• 7:34
  then from a skillset perspective.
• 7:35
  Again, the thing I love about data rolls is that all of the
• 7:39
  knowledge is out there and it's available for you.
• 7:40
  Like, you can learn python, you can learn statistics, you can
• 7:43
  learn machine learning, you can learn how models work.
• 7:46
  You can learn math, and that's, it's definitely one of the most
• 7:49
  accessible fields from an information and resource
• 7:52
  availability perspective. And so for me, I just spent a
• 7:56
  lot of time kind of learning how to code, and How infrastructure
• 8:00
  Works in learning how where to get data and how to get data and
• 8:04
  kind of went from there. So that's kind of my thought on
• 8:07
  my own career but I think for people who are thinking about
• 8:10
  Career Development, the field is very large obviously and you
• 8:14
  know you it's hard to be a specialist in all the different
• 8:17
  areas that participate in the world of data.
• 8:22
  Yeah but it's there's something for everybody.
• 8:24
  So no matter what you're sort of interested in, I think the
• 8:27
  number one thing I would always say is never to Discount like
• 8:29
  your Prior experiences because those will help you be better if
• 8:33
  you are sort of switching careers.
• 8:34
  But if you're young and you're kind of just deciding what
• 8:37
  you're doing, now then, like try different things, right?
• 8:39
  And see what you like and sometimes you like modeling a
• 8:42
  bit more than programming sometimes you really like
• 8:44
  building software and maybe machine learning Ops is more for
• 8:48
  you, right? Or data engineering versus the
• 8:49
  pure sort of data science. But there's something for
• 8:52
  everybody which I think is makes this field really great.
• 8:55
  I love that. You had a little stint in the
• 9:00
  land of Ethics because I feel like that is something that we
• 9:04
  don't see that much or even mention that much and this maybe
• 9:11
  me being totally biased but I feel like when we talk to people
• 9:16
  about certain challenges that they're having in the workplace,
• 9:21
  you rarely hear the question being asked, should we do this?
• 9:25
  It's a lot of. Can we do this?
• 9:28
  And if it's feasible, if it's possible, then we're going to do
• 9:30
  it. But we never take a step back
• 9:33
  and say, should we do this? And how are we looking at the
• 9:37
  data that's being collected? Is it being collected in quote
• 9:40
  unquote, ethical way, right? So, I really appreciate that.
• 9:45
  And just a bit of a tangent to that.
• 9:47
  I wanted to go down with you before we go, like into specific
• 9:52
  ways that you've seen different paths.
• 9:55
  And people take the different, I see path or the managerial path.
• 10:02
  I wanted to take this tangent and look at the idea of telling
• 10:07
  stories with data and it kind of came up when you were talking
• 10:11
  about it. What in your mind has helped you
• 10:16
  become better at telling stories with data?
• 10:19
  To be honest, I would say that a lot of my Consulting experience
• 10:23
  has actually made me a better communicator.
• 10:25
  Overall. So, when I was working in
• 10:27
  Consulting, there is a book called The Pyramid principle,
• 10:30
  and it's sort of the Consulting Bible for anyone.
• 10:32
  That's worked at McKenzie. At some point, we'll probably
• 10:35
  have been forced to read this, but it was recommended to me at
• 10:38
  the time that I work there, and it was really about kind of sort
• 10:40
  of structuring your thought process.
• 10:42
  But also structuring your Communication in anything from
• 10:46
  emails, to presentations to analysis that you've done.
• 10:49
  And so I've kind of taken that and I apply it now every day.
• 10:53
  So anytime that we build some sort of Of data solution.
• 10:56
  I think, you know, understanding what your who your audience is
• 10:59
  and who you're talking to. And for the most part, we're not
• 11:01
  really talking to people that are going to understand our
• 11:05
  technical jargon or infrastructure that were
• 11:07
  referencing or what data bricks means or an S3 bucket or
• 11:11
  something like that. And so these are not useful
• 11:14
  terms for us to use and so when we're explaining kind of the
• 11:17
  solutions and thinking about our objective of data storytelling,
• 11:20
  our objectives are typically that we want to share our
• 11:24
  approach and our Mechanism of arriving at some sort of collage
• 11:28
  that conclusion with somebody because often there's many
• 11:30
  different ways to look at the same problems.
• 11:32
  We want people to be aligned with the process that we took to
• 11:35
  arrive to that conclusion. And so we need to walk them in a
• 11:38
  way that they'll understand with what was the problem that were
• 11:41
  solving for. You know, here's what I thought
• 11:43
  about when I collected the data that I selected, here's sort of
• 11:47
  the models that I tried in the different approaches.
• 11:48
  I tried in this one worked best, here's why it worked best.
• 11:52
  And then based on this, here's my recommendation.
• 11:55
  And then sort of from an infrastructure perspective.
• 11:57
  We're, if we're getting into building data products and in
• 12:00
  software, the second thing we always want to do with
• 12:02
  storytelling is Garner empathy from stakeholders on why it's
• 12:06
  taking us so long to answer a question and why it's not as
• 12:09
  simple as running. A SQL query and pulling and
• 12:12
  three data points and giving them an answer, right?
• 12:14
  Because these are complex things.
• 12:17
  These are complex questions that we're trying to answer.
• 12:19
  And so being able to sort of explain in a simple way, what
• 12:23
  we're looking to build. Why our This is that we're
• 12:26
  proposing is a better solution, then a simple sort of query or
• 12:30
  something that's more scalable and especially with ML Ops and,
• 12:33
  you know, putting things into production.
• 12:35
  And again just building software all together with data products,
• 12:38
  is one sort of sum sum summation word for all of that.
• 12:42
  But these things take a lot of time.
• 12:43
  And so I think storytelling is really about communicating your
• 12:46
  objective of what? You the approach that you've
• 12:49
  decided to go with and why and why that's the best approach.
• 12:53
  And to, if it's going to take you some time, Time to build
• 12:56
  something that scalable and sustainable and can be used in
• 12:59
  an ongoing basis, being able to articulate why that is to people
• 13:02
  who don't really have an innate understanding of the
• 13:05
  infrastructure and the processes you follow.
• 13:08
  But yeah, the pyramid principle I think was great for me and now
• 13:10
  I try to kind of teach my teams as well to use that same
• 13:13
  approach of like starting at the top, like key messages, break it
• 13:17
  down into further information and then break it down into the
• 13:20
  details if people are interested.
• 13:23
  And have you found any other tricks for being the translator
• 13:29
  when it comes from taking information from the business
• 13:32
  side and bringing it to your data teams?
• 13:35
  Well, I mean, I think for data teams is equally as important to
• 13:38
  understand the business. And oftentimes data teams are so
• 13:41
  caught up in their own kind of world of working with data and
• 13:45
  working with solutions that they don't have a lot of time to
• 13:48
  spend understanding what's going on.
• 13:50
  So an example of where I use this My day-to-day is we have
• 13:55
  data warehouses that, hold a whole bunch of data, obviously
• 13:58
  about our customers and our business, which we use for a lot
• 14:01
  of our models. But that data gets generated
• 14:03
  somewhere and typically gets generated on our website or on
• 14:06
  an internal tool that were using.
• 14:07
  So what I like to do with new hires is I always walk them
• 14:10
  through where Theta comes from and where it's generated and
• 14:14
  then how that gets fed into the data warehouse and the reason I
• 14:17
  do that is because I find, it gives them an understanding of
• 14:19
  that customer journey to say, okay, when I'm looking at a
• 14:22
  couple Customer. This is what the person is doing
• 14:24
  on the website. This is how they're making a
• 14:26
  decision, or an internal person. This is the tool they're using.
• 14:30
  And when I do it that way, I found it's helpful to then
• 14:33
  articulate the problem. We're trying to solve.
• 14:34
  So if this is, you know, the customer experience and the
• 14:38
  customer let's say is looking at recipes and they have 100
• 14:41
  recipes facing them. They don't know which one to
• 14:43
  choose. This is why we need to solve the
• 14:45
  problem of a recommendation engine.
• 14:47
  For example, because we should be able to give them their kind
• 14:49
  of top five that we think they'll like so, Wind I like
• 14:53
  data origination piece I think helps humanize the process and
• 14:58
  then data people can empathize with what the user is going
• 15:01
  through and that helps them. Become better problem solvers
• 15:04
  because they're putting that user experience lens on.
• 15:07
  I really like that point around, helping your team, understand
• 15:12
  what data means. Yeah, I think that as much as
• 15:15
  business teams need to understand, you know how data
• 15:18
  empowers them to data team itself, also needs to First and
• 15:23
  you know what, how what they're doing, you know is represented
• 15:27
  in in the context of the business and the data and it's a
• 15:30
  very bi-directional flow that has to happen.
• 15:32
  I want to Pivot this for discussion slightly and talk a
• 15:35
  little bit about technical management, Engineering
• 15:38
  Management and how you got into that and how you like it?
• 15:43
  Can you start off maybe by telling me a little bit about
• 15:45
  like what your day-to-day role looks like.
• 15:47
  As a associate director at hellofresh and what's your
• 15:50
  distribution of time looks like Definitely.
• 15:54
  So, we have on my team specifically, we've got an
• 15:57
  engineering team, a bi not imagine team in a data science
• 15:59
  team. And so, we have a good mix of
• 16:01
  sort of business analysis. Type folks, as well as very
• 16:04
  technical folks that are Building Solutions.
• 16:08
  I would say that my, my day-to-day personally, I spend
• 16:11
  quite a bit of time with stakeholders to understand the
• 16:13
  challenges that they're facing. So, my team supports all of the
• 16:16
  different business departments in hellofresh Canada, so
• 16:19
  everything from marketing to product to Ops procurement.
• 16:22
  And so it's a really A wide portfolio and requires that sort
• 16:25
  of broad understanding of what the business is doing.
• 16:27
  So I think for me personally, my day-to-day is kind of
• 16:31
  understanding what people are dealing with, where we can step
• 16:34
  in and making an assessment of where it's appropriate for us to
• 16:36
  step in versus having a stakeholder.
• 16:39
  Do something themselves or maybe manage their own solution and
• 16:43
  try to kind of assign value value to all of the projects we
• 16:47
  do and values hard to measure, right?
• 16:48
  Because you're investing a lot of time and building.
• 16:51
  Again, either a software or A model and deploying that
• 16:54
  somewhere and maintaining that and building it into processes.
• 16:57
  You're spending a lot of time doing that.
• 16:58
  So there has to be some sort of value.
• 17:01
  And so from a business perspective, I'm always thinking
• 17:03
  about Revenue generation versus cost production cost.
• 17:05
  Production is always easier to measure value on because you
• 17:09
  typically know how much something costs and now you're
• 17:11
  trying to sort of reduce that but with Revenue generation it's
• 17:14
  hard because we can never fully attribute any sort of Revenue to
• 17:18
  any sort of data solution. The Data Solutions will
• 17:21
  certainly help. Help.
• 17:23
  But we don't know exactly how much revenue can be attributed
• 17:26
  there. And so most of the time my
• 17:28
  strategy at least, when I initially started building, this
• 17:31
  team was to kind of build trust in data among stakeholders.
• 17:34
  So delivering value very quickly for a period of time so that
• 17:38
  they can see that there's value in letting data teams.
• 17:41
  Take the time to build something more complex, I think is the
• 17:44
  leader of a data team is as a manager, your job is to buy your
• 17:47
  team time. And the way you by time is by
• 17:50
  prioritizing some quick to value.
• 17:53
  First, so that your team has time to work on those more
• 17:55
  complex things that, you know, will add value but may not be so
• 17:58
  apparent the business stakeholders immediately in
• 18:02
  terms of my team. So every day to day for them
• 18:05
  there, you know, in in the weeds of building a lot of our
• 18:08
  Solutions. So whether that's models whether
• 18:11
  that's, you know, even software again.
• 18:13
  So we build a number of different softwares to enable
• 18:16
  access to data, or to provide is really data products that were
• 18:19
  building. So they're really in the weeds
• 18:20
  of that Pipelines. In our databases are data
• 18:24
  storage. So they're really in the weed in
• 18:27
  the weeds of all of that. And I think the one interesting
• 18:29
  thing at least that I found as sort of becoming a manager of a
• 18:32
  technical team is that your team will always know more than you.
• 18:35
  So there's always that sort of you cut like what's your role on
• 18:39
  the team as a manager is to buy them time so they can focus and
• 18:43
  to continue to build trust in data amongst your stakeholders
• 18:47
  and to find the right opportunities and prioritize
• 18:49
  where they should be adding time but that the sort of trade-off
• 18:52
  there You don't have that much time yourself to keep up with
• 18:55
  the technology and to keep up with what everybody is doing all
• 18:58
  the time. And so, I think when people are
• 19:00
  thinking about that sort of management versus individual
• 19:03
  contributor path, you know, it really depends on your personal
• 19:06
  interests. And if you are somebody that
• 19:08
  wants to be really involved, and it's in the tools and Building
• 19:12
  Solutions, and actively Building Solutions, and all of that
• 19:15
  stuff, then you should be an individual contributor.
• 19:18
  And as soon as you become a manager, you're now spread a
• 19:21
  little bit thinner. Or you know opining on a lot
• 19:24
  more things, your you become more of a translator and
• 19:27
  identifying requirements determining what the team is
• 19:30
  doing. Helping them think through the
• 19:32
  flow thinking about scalability. So if we build a solution now
• 19:36
  how is that going to change in two months, three months, eight
• 19:39
  months, with the business strategy.
• 19:40
  So if the business grows my going to need to be ready to
• 19:43
  add, something to the solution or change, the way that it
• 19:45
  works. Whereas your team on the ground
• 19:47
  isn't going to be thinking about that because they're just
• 19:49
  thinking about getting the one sort of thing done.
• 19:52
  So, I think that's sort of how the rules change, but it's an
• 19:56
  interesting experience. And, you know, I often find
• 19:58
  myself feeling that pressure of like, how do I keep up?
• 20:02
  And I need to know all the things and my team members will
• 20:04
  find new things that they're using all the time.
• 20:06
  And sometimes I'm like, I've no idea what this is, and but it's
• 20:10
  not and I get to learn all the time and I learn from them.
• 20:12
  And, you know, at the same time, learning to let go a little bit
• 20:15
  and not control everything. Yeah.
• 20:18
  You know, I think there's so much great that you good stuff
• 20:21
  that you said there and particularly, Sure, that point
• 20:23
  around how your team will always know more than you write and how
• 20:27
  you have to. You know, Elevate the way that
• 20:32
  you were to a different level in order to, you know, really take
• 20:36
  advantage of their talents best and their time in the best
• 20:38
  possible way. Was it difficult for you to give
• 20:41
  up day-to-day individual contributing work and sort of
• 20:43
  coding? Yeah.
• 20:44
  I mean at first. Yeah, because you you're used to
• 20:47
  controlling how something works and you have your own way of
• 20:50
  doing things. And other people are going to
• 20:52
  have their own way of doing things or their own thoughts.
• 20:55
  I think at first, it's hard to not have that control.
• 21:00
  Is your used to sort of controlling and but I think it's
• 21:02
  important to as a manager. And as a leader, you have to
• 21:06
  sort of figure out what is your role within your team.
• 21:08
  So you have a team of 15, 20 people, let's say, and they all
• 21:11
  work together and everybody has a role and your role isn't to at
• 21:15
  least in my view. Your role isn't to be managing
• 21:18
  people to a tee and telling them exactly how to spend their time
• 21:21
  or exactly what to do, your role is to be training them, teaching
• 21:24
  them best practices, but that's what those are there, best
• 21:27
  practices, you're not kind of like Rolling every single
• 21:30
  element. So I think for me initially it's
• 21:33
  always hard for everybody to give up that because you're
• 21:36
  ultimately accountable for anything that your team does.
• 21:38
  So you obviously need to have some oversight and an
• 21:41
  understanding of what they're doing but then also your role
• 21:45
  changes. And so for me it was about kind
• 21:46
  of recognizing that in my new role, the value I add to my team
• 21:50
  isn't duplicating their work or you know working with them on
• 21:54
  something or telling them how to do something to Value.
• 21:56
  I add is helping them navigate the complexity of A company that
• 22:00
  we work in helping them navigate and build a solution that's
• 22:03
  going to be useful and then helping them sort of prioritize
• 22:06
  what to build first and how to think about the long term that
• 22:09
  said though. We are my team at least like I'm
• 22:11
  still pretty Hands-On and I do work on my own projects as well.
• 22:15
  So sometimes if it's something where you know, I think
• 22:18
  especially if I'm just scoping something out.
• 22:19
  Like I usually do that myself before I give it to anybody on
• 22:22
  the team because I don't know yet if I did and again, that
• 22:25
  role of identifying of something as valuable or not, sometimes I
• 22:28
  don't know. Yet if something Is doable or if
• 22:30
  it's valuable. So, I might build like a quick
• 22:32
  prototype myself, just to kind of see if it's something worth
• 22:36
  pursuing. And then once I'm confident,
• 22:38
  I'll give it to someone on the team that can do at times times
• 22:40
  better than me. And then, they'll sort of take
• 22:43
  care of it from there. So I think there is always still
• 22:45
  time for you to do what you enjoy and for myself to do what
• 22:48
  I enjoy. But that time becomes less.
• 22:51
  So am I going to be building, you know, complex software that
• 22:55
  is truly scalable and in production and stuff like that,
• 22:58
  probably not but I can still do a bit of that upfront work and
• 23:02
  work with the team to get through to those end goals.
• 23:06
  I think that what you really point out with your responses is
• 23:09
  that the context in which you work is as important as what you
• 23:13
  work on. Yeah.
• 23:14
  And I think in many ways that represents the, the trade-off
• 23:18
  between you don't management and contribution, which is really
• 23:21
  something everybody does, you know, even if you're a, you
• 23:24
  know, a junior ic50 person team, you're still managing up, you're
• 23:29
  still working. With people to help them
• 23:31
  understand what it is. You're actually doing and be
• 23:34
  part of the reason I wanted to talk about this was because it's
• 23:36
  an active discussion. I think in our community many
• 23:40
  times you know for many members like do I go down the management
• 23:43
  track and maybe spend some times engineering manager or do I
• 23:46
  spend more time you know on my individual contributor piece and
• 23:49
  then kind of go down. Maybe the tech lead side or the
• 23:51
  senior to staff. And you know I think we're
• 23:54
  actively in a moment right now and Engineering world where
• 23:56
  we're starting to think more about what does?
• 23:59
  Good leadership for technical work.
• 24:01
  Look like as the importance of technical work to business value
• 24:06
  has grown. Yeah, right.
• 24:08
  Yeah, it is. And it is, I think it's a topic
• 24:11
  amongst anybody that works in this space even people that have
• 24:15
  been doing it for some time because, you know, the types of
• 24:17
  things that I think about is, you know, motivating the team.
• 24:21
  And right now we have a huge Talent Gap in Tech, like, people
• 24:24
  are. We know this to be true, people
• 24:26
  are always moving around and if you think about, what does it
• 24:28
  take for me? To get value out of somebody.
• 24:31
  So do I really in traditional roles like outside of tech for a
• 24:34
  long time when I was working like you'd probably spend, you
• 24:36
  know, two years in a job, 23 years in a job and that you
• 24:39
  know, the first six months were sort of like onboarding and then
• 24:41
  you kind of add value for six months and then you maybe start
• 24:45
  thinking about your sort of next move.
• 24:46
  But the world is not like that now so it could sort of begs.
• 24:49
  This question of as a technical leader one, how do you get value
• 24:54
  from people very quickly and you're not going to do that by
• 24:56
  being very like micromanaging? Because people want to work in a
• 25:01
  place where they have freedom and autonomy to build the
• 25:04
  Solutions in the way that they want to build them and they in a
• 25:07
  bit, you know, with attack. It's a bit of a creative field
• 25:10
  as well because there's many ways that we can solve the same
• 25:12
  problem. So, giving them that autonomy is
• 25:14
  something that people probably value in this type of space.
• 25:17
  But then at the same time, you have to, as a leader balance
• 25:20
  that autonomy and what people are building with the long-term
• 25:23
  Vision, which is a little bit more difficult, because as
• 25:25
  especially in a case of, you know, hellofresh our business
• 25:28
  moves very quickly and things change.
• 25:29
  Very quickly. And so we build some, you know,
• 25:31
  solution model, software whatever.
• 25:33
  Today, in 6 months, it's going to be different because we're
• 25:36
  going to have different products, different parts of our
• 25:37
  business. And so that needs to evolve.
• 25:39
  So how do we build something products and ways in a way that
• 25:43
  they're going to be scalable? And so, that if that person
• 25:45
  leaves in a year, somebody else can just sort of pick it up and
• 25:49
  run with it or continue to build on top of it.
• 25:51
  And I think, and then that's where we kind of get into this
• 25:54
  sort of element of like governance, right?
• 25:57
  And and really Tech governance and data beta product
• 25:59
  management. Tech product management at the
• 26:01
  importance of, not only defining requirements of what techniques
• 26:05
  to look like. But also defining long-term road
• 26:08
  maps that are really attached to business strategy and designing
• 26:12
  Tech in a way that's going to be scalable documenting, how things
• 26:15
  work. Having a really common
• 26:17
  understanding. I mean if you even think about
• 26:19
  anybody operating in an AWS environment, there's over 600
• 26:23
  services in AWS which means that anyone developer can take a very
• 26:28
  different path to how They build a solution and not everyone's
• 26:32
  going to understand all the services so it's not as simple
• 26:34
  as just handing this over to a new person that starts on your
• 26:37
  team. So really taking that time to
• 26:39
  like document and tell that story of like, here's what we
• 26:42
  built. Here's why we build it this way.
• 26:44
  Long-term considerations, if it grows, if it needs to scale.
• 26:47
  Look, here's suggestions on how you might navigate that and it's
• 26:50
  almost like managing the turnover that we know is going
• 26:53
  to come so that we can get value from people in the time that
• 26:57
  they're with us, we can make sure they're working on
• 26:59
  something. You know, hopefully they stay.
• 27:00
  But we can make sure they're working on something that is
• 27:03
  valuable for them, and valuable for us and then we're sort of
• 27:05
  building for that future to for the longer-term, feel like I
• 27:09
  went on a bit, sort of a different answer there, but no,
• 27:12
  no, I think these are, these are honest Reflections from your
• 27:15
  time, as a manager and what you found successful about
• 27:20
  motivating a team, getting the most out of them which is, which
• 27:23
  is really crucial, you know, I think there's a heavy emphasis
• 27:26
  on hiring the best talent in the discourse.
• 27:29
  Not the heaviest emphasis on getting the most out of the
• 27:32
  talent when it walks in the door, which is a much harder
• 27:34
  problem, right? And that is, honestly, what's
• 27:39
  more realistic for a lot of companies, right?
• 27:41
  It's get the most out of what you have, not sit there and pine
• 27:43
  for something better, you know. Yeah.
• 27:47
  And so, a Mendham just spending so much time and money.
• 27:51
  If you do, give someone that maybe isn't a good fit or
• 27:54
  they're just such an incredible talent that they hop jobs.
• 27:59
  Once every six months like we're saying and which is a real
• 28:03
  thing. Yeah.
• 28:04
  And actually when you were talking about that, it reminded
• 28:07
  me of this song that I wrote is not really a full-fledged song,
• 28:11
  it's just a few chords on the guitar and it was I played it,
• 28:15
  did he, and I threw it on Tick-Tock in case Vision.
• 28:18
  I want to see it. You can do it with me.
• 28:20
  Vishnu. Nobody knows this about Vishnu,
• 28:22
  but he is, he was like a professional singer.
• 28:26
  Well, I'm waiting for you to deal.
• 28:27
  Yeah. For some Duets Vishnu.
• 28:32
  I'm gonna have to make a Tick-Tock account, just for
• 28:34
  this. So, but anyway, yeah, it's like
• 28:38
  you joined, like, subscribe. The the whole thing of the song,
• 28:43
  like the lyrics are, this is the story of the ml engineer who
• 28:46
  changes their job once or twice a year.
• 28:50
  Technical debt is building up so they gotta get the fuck out of
• 28:54
  here. Oh, I love that.
• 28:59
  Yeah. Outside Melody to it, you know,
• 29:02
  and I like that and I think that really resonates with me because
• 29:07
  I think this is when people kind of look at Career Development
• 29:10
  to, you know, a lot of people like to build things me
• 29:13
  included. And I like to, you know, one of
• 29:15
  the things I love is like standing up a team and building
• 29:17
  up a team from scratch. And so, I've tripled the size of
• 29:20
  my team since I started here last year and part of it is
• 29:23
  because I like taking sort of this blank space and starting
• 29:27
  from nothing and sort of going up something.
• 29:30
  But then there comes a point where now you have to like
• 29:33
  maintain the team and grow the complexity of what it's doing.
• 29:37
  And so they're sort of I think when it comes at least the team
• 29:40
  growth into management, you sort of start with building a team
• 29:43
  and then managing the complexity as that team grows managing the
• 29:48
  sort of norming stage of the team, where the operations have
• 29:52
  to become a little bit more normal.
• 29:53
  But I think was with software and any of the complex solutions
• 29:56
  that individuals, bill, that's kind of the same thing because a
• 29:59
  lot of people Really like to build things.
• 30:00
  Not a lot of bull like to maintain things and manage
• 30:03
  complexity. And so the both are very like
• 30:05
  critical experiences to get. And so these people, the people
• 30:08
  that generally tend to jump around, they come into a
• 30:10
  company, they build up something really great.
• 30:12
  Then we got a lot of that technical debt for whatever
• 30:15
  reason, and then somebody else has to sort of come in and
• 30:17
  either clean it up in a lot of times, they don't want to clean
• 30:20
  it up. They just bought something new
• 30:21
  because that's easier and so you have all these Builders but
• 30:24
  yeah, and I think it's an important consideration for team
• 30:27
  management of. How do you get people that are
• 30:28
  Builders? But also So people who have that
• 30:30
  in-depth understanding of the tack and can maintain an
• 30:33
  evolved, something like take something that has existed for
• 30:37
  some time and make it better versus just building something
• 30:40
  new and putting you in the same position.
• 30:43
  Yeah. Especially when they're not the
• 30:44
  ones that built it. Like you said, it is a huge
• 30:47
  point that people I think are a little bit less excited about
• 30:52
  fixing something that they didn't build.
• 30:54
  And they have to go read the documentation or they have to go
• 30:57
  figure out like how it was set up and why.
• 30:59
  And you spend a lot of time and energy, just trying to navigate
• 31:03
  The Labyrinth. But I've got kind of, I don't
• 31:07
  know if I would call it a controversial question, but I
• 31:09
  would love to hear your thoughts on this.
• 31:11
  How much do you feel growing? Your team actually grows your
• 31:17
  productivity. Oh, not much.
• 31:19
  I feel like I would rather have a smaller team that's highly
• 31:22
  productive than a bigger team. That is not as productive.
• 31:25
  And I think, at least my view is I sort of think of it as a It'll
• 31:29
  SWAT team. And it's because if you're, if a
• 31:32
  company is investing in a data team, any version of a data
• 31:34
  team, we know it's going to be expensive, we know it's a call
• 31:36
  center and so you really need that team to be delivering
• 31:39
  value. I think when you manage a team
• 31:41
  that has a lot of resources, they start to do a lot of low
• 31:43
  value work, because business partners will Handover sort of
• 31:47
  that little value work, right, like reporting running, SQL
• 31:50
  queries, a bunch of social, then you're gonna have a whole bunch
• 31:52
  of people working on that when their talons are probably better
• 31:55
  used doing something else. But if you have a sort of
• 31:58
  smaller team, That you can protect you've proven the value
• 32:02
  that the ad, you can protect their time, you can have them
• 32:04
  focus on high impact work. It works for both parties
• 32:07
  because then you can, you know, they're happy because they get
• 32:10
  to kind of kick out projects probably very quickly and they
• 32:13
  can do a whole bunch of different things and they get
• 32:17
  that sense of like importance and contribution as well.
• 32:19
  So they know the value that they add to an organization and at
• 32:23
  the same time, you keep business partners at Bay a little bit
• 32:26
  because they know that your team's time is valuable.
• 32:29
  And so if Are going to give you something, it better be
• 32:31
  something needy and have better be like a problem that you're
• 32:33
  solving. And so like to me that's sort of
• 32:36
  the approach and that I think works best.
• 32:38
  Obviously the size of the team should be sort of in
• 32:41
  correspondence with whatever the size of the company is.
• 32:43
  But yeah I'm all for small Mighty team versus something
• 32:49
  bigger that you kind of lose track of where people are
• 32:52
  spending their time and it may not be that valuable.
• 32:55
  Hi. This is my Alaric.
• 32:57
  I'm the founder of pun intended Innovation and Betty I and this
• 33:01
  is a quick Public Service Announcement.
• 33:02
  If you're tired of having nightmares about the machine
• 33:05
  learning models, you have in production, then I encourage you
• 33:07
  to subscribe to the ml Ops Community podcast, it's where
• 33:11
  all the cool ml Ops kids hang out and learn how to tame their
• 33:14
  ml models. Yeah, the reason I ask is
• 33:17
  because I just think about how a lot of times you're almost
• 33:22
  pressured to grow and it it looks great on your resume if
• 33:25
  you manage 20, people or 40 people, or the more people that
• 33:29
  You managed the better, it looks especially if you're a manager
• 33:33
  and you're taking that managerial path and I think
• 33:36
  about how just in the community like not, I don't even manage
• 33:41
  really but just the more people sometimes that come into the
• 33:45
  community and there's always a lot of people that are willing
• 33:48
  to help and I love that. It's so cool to see.
• 33:51
  But sometimes there's that saying where it's like too many
• 33:55
  chefs in the kitchen, spoil the soup.
• 33:57
  Yeah. And it ends up being Like yeah,
• 34:00
  actually, like I would love for you to help out but you know
• 34:03
  what, like I don't know if having more people on this
• 34:07
  problem is going to solve the problem.
• 34:09
  It's only going to make the problem more like organizational
• 34:14
  mess. Yeah.
• 34:16
  And I think on your first comment about managers striving
• 34:19
  for bigger teams. I think that's just the
• 34:21
  fundamentally wrong way and how we measure success.
• 34:24
  I think, you know, your role as a manager is to.
• 34:28
  Again, find the Things that are most pressing for a business and
• 34:31
  to help your team. Prioritize them and give them
• 34:33
  guidance on how to build those Solutions.
• 34:35
  And so, your, your performance should really be measured on
• 34:38
  effectiveness of what you've delivered and Effectiveness.
• 34:41
  In allocating the team's time to the most pressing needs.
• 34:44
  But, yeah, I totally hear you. Also the, that's the other sort
• 34:47
  of Challenge on the bigger team because you're going to have
• 34:50
  different personalities, different people that solve
• 34:53
  problems very differently, people that have, you know, want
• 34:56
  to opine on different parts of a problem or You know, pissing on
• 35:01
  the hydrant, it's what I used to call it back in vain.
• 35:03
  Oh my God, that's so true. Yeah, I like a lot of people
• 35:06
  want to focus on the hydrogen and, you know, and that's fine.
• 35:09
  Like people, everybody's going to have an opinion about
• 35:12
  something and, you know, it's almost like the work gets really
• 35:15
  diluted. So when you have sort of one
• 35:17
  meaty thing, a lot of people want to take part in it because
• 35:19
  they all want to have that as part of their own personal sort
• 35:22
  of accomplishments versus in an environment where you have lots
• 35:26
  of work and fewer people than you, probably probably not going
• 35:30
  to worry about that too much because they're always be enough
• 35:32
  for everybody and people will more likely to focus on their
• 35:35
  own specific thing. But yeah, it's an interesting
• 35:38
  problem. I think this is a great
• 35:42
  discussion and I would love to add context to your by talking
• 35:46
  more about what hellofresh actually is, what your team, you
• 35:51
  know, sort of does in terms of actual like, sort of product
• 35:54
  development, Etc. I know all of us.
• 35:57
  Every single one of us at least the u.s.
• 35:59
  Has gotten one of the little slips.
• 36:01
  Okay, that gives us some X number of dollars off of 10, 10
• 36:05
  meals. So I know we're your meal kit
• 36:07
  company, but could you give the listeners a little bit more of a
• 36:10
  sense of like, how like, what hellofresh actually is?
• 36:14
  Yes, that point. Absolutely.
• 36:16
  So I think we're a meal solution companies where a food solution
• 36:19
  company and I would say, from a data perspective this is
• 36:22
  definitely the most interesting business I have ever worked at.
• 36:25
  I think it's a extremely interesting business model
• 36:28
  because it has all the all the Challenges that any business
• 36:31
  might have plus more. So basically the way our
• 36:34
  business works is you can go on the app or on the website, you
• 36:37
  can select a bunch of meals that you'd like to eat later in the
• 36:41
  week. We send you boxes pre-portioned
• 36:44
  ingredients for every single meal and then you cook your
• 36:47
  meals at home in the u.s. we also have a few other offerings.
• 36:51
  We have ready to heat meals for examples if you don't want to
• 36:53
  cook, but you just want to heat up your meal, then it's
• 36:55
  precooked for you. So, we've got the factor brand
• 36:57
  in the US which does ready to heat.
• 36:59
  I'm in a few smaller brands with that specific diet.
• 37:02
  So if you want like vegetarian only we have something called
• 37:05
  Green Chef. I think that's in the US we have
• 37:08
  every plate which is more of a value version of our meal.
• 37:10
  So we've got a few different brands but in the Crux of it our
• 37:13
  role is to create meal Solutions and Food Solutions.
• 37:18
  And it's it's interesting, right?
• 37:19
  Because we're so on the business model side, this is really
• 37:22
  interesting because we're not only competing with grocery
• 37:27
  stores but were actually competing with this habit, that
• 37:29
  people. Have and the habit of going
• 37:30
  grocery shopping every week and then coming home and cooking
• 37:33
  meals and meal planning. So we're trying to take that
• 37:36
  away that sort of activity of having to meal plan.
• 37:39
  We know people are busy. If you, you know, tend to order
• 37:43
  out, then our food is for you. So if you're spending a lot of
• 37:46
  money ordering overeats or whatever you're doing, this is
• 37:48
  much better for you. You can make it yourself.
• 37:50
  It's cheaper. And so it gives you that
• 37:53
  convenience factor without, yeah, just gives you a
• 37:56
  convenience factor without those sort of additional costs and all
• 37:58
  that stuff from from Uber. So I would say our target market
• 38:01
  is people that are looking for convenience, but from a business
• 38:05
  model perspective, why it's such an interesting data problem is
• 38:08
  because we've got the e-commerce, which is obviously
• 38:10
  people ordering and then sending food to them.
• 38:13
  We also produce all of the boxes ourselves.
• 38:15
  So we have a huge kind of distribution facility where we
• 38:18
  package, all of the boxes. We have a Pick and Pack
• 38:20
  operation in Canada. We actually have a manufacturing
• 38:22
  operation as well where we manufacture some of our own like
• 38:25
  sauces and things like that. And then we have Sort of that
• 38:29
  production planning and then our own sort of logistics that we
• 38:32
  deliver boxes. We also have our own chefs and
• 38:35
  culinary teams that Designer menus, would you all over ohm,
• 38:38
  like photography and user experience stuff ourselves as
• 38:42
  well. So, from a data perspective,
• 38:43
  it's interesting because you have the complexity of being
• 38:46
  kind of a just-in-time operation, where you want to
• 38:50
  have the right recipes, displayed to the right, people
• 38:53
  at the right time so that they can make their choices and have
• 38:57
  enough choice, and you want to have sufficient choice for them.
• 38:59
  So they don't Feel like they're boxed into anything and that's
• 39:02
  sort of on the revenue side and then on the cost side, you want
• 39:05
  to have the right raw ingredients at the right time in
• 39:08
  your facility so that you can produce the number of boxes.
• 39:11
  You need with at the same time minimizing waste and not
• 39:14
  creating any any food waste which is one of our missions on
• 39:16
  sustainability We Exist because we want to eliminate food waste
• 39:20
  and then toward of delivering those boxes to people on time
• 39:23
  and keeping them cold right because we're delivering
• 39:25
  perishable food items. So there's a big time element
• 39:28
  and also with perishability Because food goes bad.
• 39:30
  So we can't hold a lot of inventory, right?
• 39:32
  We can't we don't have that availability ability like an
• 39:35
  Amazon to have a huge distribution warehouse with a
• 39:38
  lot of inventory. We have to operate in that
• 39:40
  just-in-time model. So it's really, really complex.
• 39:43
  I think anyone who's, like, ever done operations analytics or
• 39:47
  optimization modeling if you want to whip out your calculus
• 39:50
  textbook, it's a lot of that kind of stuff, so it's a perfect
• 39:54
  nice perfect company for that. But yeah, it's super
• 39:57
  interesting. Yeah.
• 39:59
  I mean it sounds like what I'm hearing.
• 40:01
  There are challenges of e-commerce cpg Logistics and
• 40:07
  grocery all combined into one a highly competitive market where
• 40:13
  margins are like, really important.
• 40:16
  Yeah, to to maintaining competitive position.
• 40:19
  Yeah, it's a volume game. Absolutely, it's a marketer in
• 40:21
  volume game and and it's an expansion game.
• 40:24
  Right. So the business has to expand.
• 40:26
  I think we're in 17 countries now.
• 40:28
  So we've expanded Get over there and we've only been around for
• 40:30
  ten years so it's huge growth for overall, but then the other
• 40:35
  the other sort of interesting thing so in Europe, where our
• 40:38
  main office is in Berlin. And so most of our operations
• 40:40
  are in Berlin. But obviously, we have Canada
• 40:42
  and the US and our challenges are also very different because
• 40:45
  Canada and the US are very big countries and so you can't get
• 40:49
  when it comes to supply chain. And you think about, you know,
• 40:51
  supply chain optimization. You know, we can't always get
• 40:55
  the same food from every supplier.
• 40:57
  Sometimes, we're going an extra country.
• 40:58
  You're two down south to get food, but that's a really,
• 41:02
  really big distance of thousands and thousands of miles.
• 41:05
  And then again, from a production perspective, if we
• 41:07
  sort of have spread out facilities, how do you optimize,
• 41:10
  which facility builds, which box for which customer and not, and
• 41:13
  how does that get delivered in the most optimal way?
• 41:15
  So that you can limit the amount of time that food is aging
• 41:18
  before it even gets to the customer.
• 41:20
  So it's really, really complex and even more complex and North
• 41:23
  America. Yeah, it's just overwhelming.
• 41:26
  This listening Demetrius. I know you have some, yeah, it's
• 41:28
  Listing to one thing that I think about with that is
• 41:32
  probably a lot of cultural differences between the
• 41:35
  different countries that you go into and I think about how my
• 41:39
  own. So I live in Europe right now
• 41:41
  and my wife is Spanish and it is very uncommon for her especially
• 41:46
  to think about eating out more than once a week.
• 41:51
  And then you go to the US and especially in the tech fields
• 41:56
  that you. You have people who basically
• 41:58
  Ali live off of ubereats and that's their like, that's their
• 42:02
  diet, specifically Vishnu. So yeah, I just think about that
• 42:14
  and how different it is in the way that people think about
• 42:17
  that. And then you, you also add on
• 42:20
  all these other very minut differences that you wouldn't
• 42:26
  really think about, but you probably are hitting Head on
• 42:29
  when you are conducting business in the different areas.
• 42:33
  So it's awesome and it's probably a good transition to
• 42:37
  talking a little bit about data engineering because yeah.
• 42:40
  So this is like you like how I did that day, but I am, I try
• 42:48
  sometimes, sometimes I get it right.
• 42:50
  But yeah, so from a data engineering perspective, right?
• 42:52
  Like we think about. So how do we, how do we learn
• 42:54
  about our customers? And how do we meet all of these
• 42:57
  differences? And so to your point there These
• 42:59
  cultural differences around eating out but there's also
• 43:01
  cultural differences around how people create food and people
• 43:04
  with big families like they're not cooking, you know, a meal
• 43:08
  for themselves and their two kids and their husband, like,
• 43:10
  they're cooking a meal for themselves and their cousins and
• 43:14
  aunts and uncles are going to come over later and have food.
• 43:16
  And so, there's always food in the house that I grew up in a
• 43:18
  household like that too. And so, this isn't work for that
• 43:21
  environment, because we're not creating batch meals.
• 43:24
  We're really trying to solve the you don't have anything to eat
• 43:27
  for dinner tonight. Probably going to order Uber eat
• 43:30
  anyway. So here's something that's a
• 43:31
  little bit healthier for you that you can cook at home and so
• 43:35
  finding that information about people so this is you know like
• 43:39
  let's say a lot of your like European cultures I grew up in
• 43:42
  Europe as well and that's how it is there but is it fair for me
• 43:45
  to assume that anybody who's European would have that
• 43:48
  lifestyle you know probably not. That's that's a completely
• 43:51
  biased and unethical assumption and so when it comes to data
• 43:55
  engineering sure we can probably buy some data about
• 43:58
  Demographics. And and there's a company in
• 44:00
  Canada at least call Denver Onyx and they provide postal code
• 44:03
  level demographic data where we can get down to the ZIP code.
• 44:06
  We can see it like generally the people that live there from the
• 44:11
  cultures, then the backgrounds that they have, and probably
• 44:14
  where they like to shop and things like that.
• 44:16
  But that's not from a feature perspective in this sort of
• 44:20
  using features in modeling, like that's good.
• 44:22
  But we're still not getting to that individual person level.
• 44:26
  So now the problem becomes. Okay, the people who Product is
• 44:29
  good for we can assume they've tried our product or will at
• 44:32
  some point by it because they already care about convenience.
• 44:34
  They want a couple of meals, they don't want food, waste,
• 44:36
  whatever. So the we've already got them.
• 44:38
  But how do we reach these other people and how do we learn about
• 44:42
  about them? How do we collect data or impute
• 44:45
  data in the, in ways that isn't presumptuous?
• 44:48
  And that we're not making these sort of broad assumptions based
• 44:51
  on personal characteristics but we're actually looking at people
• 44:54
  as individuals and seeing the lifestyles that they like to
• 44:57
  lead and and that's really A tricky thing and I think I did a
• 45:00
  presentation a little while ago, but one of the things I always
• 45:03
  talk about is like manual data collection versus on like
• 45:06
  automatically generated data. And I think a lot of social
• 45:10
  media companies, for example, or Internet native companies, they
• 45:13
  don't need to collect a lot of data about you.
• 45:15
  So like a Facebook, for example, you know, if you don't tell them
• 45:18
  how old you are it doesn't matter.
• 45:19
  They'll figure out how old you are because they have enough of
• 45:21
  your engagement data so they'll collect that about you but we
• 45:25
  don't really have access to that.
• 45:26
  And so we do have to depend a lot on.
• 45:28
  Those traditional data collection methods and you know,
• 45:31
  the world of data science and statistics.
• 45:34
  That used to be the truth before big data that you'd go out there
• 45:36
  and you would figure out how to sample population.
• 45:39
  You'd figure out what questions to ask.
• 45:41
  You'd go through the whole kind of survey design process and we
• 45:44
  find ourselves having to do that.
• 45:45
  So that we can collect data about people that we haven't
• 45:49
  reached yet to understand how to design or product in a way
• 45:52
  that's going to be relevant for others as well.
• 45:55
  Yeah, the bit of a tangent there but Well, it makes sense to
• 46:00
  that. You have thought about this, and
• 46:02
  it's nice. I'm sure that you have that
• 46:04
  ethics background and that side of things because like you said,
• 46:10
  don't never discount your past and how it can help you in the
• 46:15
  present. And I see that shining through
• 46:18
  as you're looking at these problems and you're thinking
• 46:20
  about these different challenges that you face.
• 46:24
  Now, let's go into a bit more of a rapid-fire session.
• 46:28
  Finish it off. I want to ask, let's start it
• 46:31
  off with a relatively simple. One, the last book you read.
• 46:36
  Oh boy. I've read a chapter from
• 46:40
  Epictetus and Nikita's discourses.
• 46:44
  It was really good. It was about.
• 46:46
  Yeah, sometimes I like to dabble into some story philosophy and
• 46:50
  remind myself not to take myself too seriously and Dad, it's all
• 46:54
  good. But anyway, just this particular
• 46:58
  one was about sort of knowing your strengths, this particular
• 47:01
  chapter was a little bit about Hercules and not knowing your
• 47:04
  strengths and so you're put in a position to No that
• 47:07
  strengthened. So the message was that, you
• 47:09
  know, Hercules wouldn't have known that he was Hercules until
• 47:12
  he became Hercules and so you don't know what you can do until
• 47:15
  you kind of put yourself in that position.
• 47:17
  So why not? I guess that's a nice connection
• 47:19
  back to my career development Shadow.
• 47:21
  Just give it a shot. You never know where it may
• 47:25
  lead. Give me Griffin gets credit for
• 47:28
  that one. I love it.
• 47:31
  Yeah that really is that was really nice.
• 47:33
  It gave me goosebumps ended. It reminded me of something I've
• 47:36
  heard. About how it's like you're never
• 47:38
  served up anything in life that you are not able to handle.
• 47:43
  So if you're going through something, you can rest assured
• 47:48
  that you are able to handle it because you're not going to be
• 47:51
  served up that thing. Unless you are at that
• 47:56
  particular point in your life that you can go through it.
• 47:59
  And so now, next one, probably a little less philosophic.
• 48:06
  It's more actually. Let no, I'll go with this.
• 48:09
  Okay, what has been normally we ask what the last bug that you
• 48:15
  smashed would be but being is that you're in the managerial
• 48:19
  position? What has been one of the harder
• 48:22
  conversations you've had to have with someone and how did you
• 48:25
  navigate that? Yeah, good question.
• 48:31
  Actually, I had to cut have a conversation with somebody
• 48:34
  couldn't doesn't really like writing documentation.
• 48:36
  Ation and he felt something really complex and I get a
• 48:39
  documentation sex and I honestly, I wouldn't say that
• 48:41
  this was card. This is I love my team and I
• 48:43
  love working with everybody on my team and it's kind of like, I
• 48:47
  know people don't like to do that stuff because it's much
• 48:50
  more boring than the actual solutioning.
• 48:52
  So it was hard because I have to now make somebody do something
• 48:56
  that I know, they really don't want to do, and that's not
• 48:58
  really my style. I like to let people be free.
• 49:02
  But anyway, we kind of have a conversation.
• 49:04
  I would say that it went well, because we just This is someone
• 49:07
  that can is very reasonable person and once we sort of
• 49:10
  articulate, the fear that if nobody else knows how the
• 49:13
  solution works, then, you know, we're going to be kind of
• 49:17
  screwed and then somebody else is going to have to rebuild it.
• 49:19
  So he understood all good now all fixed, but yeah, but it's
• 49:24
  hard. I think that's the hardest thing
• 49:25
  for me is sometimes like making my team do things, they don't
• 49:28
  really want to do. Because I really try to create
• 49:29
  an environment where people are, you know, enjoying the work that
• 49:32
  they're doing. I find we get the best outcomes
• 49:34
  from folks when they They enjoy the work that they're doing.
• 49:38
  So I have to try to think of like creative ways to them to
• 49:41
  encourage them, and motivate them to do these less fun
• 49:43
  things, but that are pretty critical.
• 49:45
  And so if I kind of relate it back to Career Development and
• 49:47
  Leadership capabilities and that usually does the trick besides
• 49:51
  the part on Career Development and Leadership capability.
• 49:55
  There's a lot of parallels with me and trying to get my daughter
• 49:59
  to do things. She doesn't want to do, like,
• 50:02
  brush your teeth at night, and I have to find fun ways to Sure,
• 50:07
  actually do that. Yeah, that's a tough one.
• 50:12
  So the next one is, what kind of marketing have you seen from a
• 50:19
  data product that made you? Roll your eyes.
• 50:22
  Oh boy. That's a good question.
• 50:29
  Honestly I can't think of many examples right now.
• 50:32
  I think mostly because I'm not really on social media.
• 50:34
  So I just avoid a lot of the that marketing that marketing
• 50:40
  that comes my way. Yeah, yeah, I don't have a good
• 50:42
  answer to that. I generally try to avoid avoid
• 50:45
  it now, it's all downhill because you're joining Tick-Tock
• 50:49
  to hear Vishnu things but that's the only Channel I'm going to
• 50:53
  follow. Is it call the channel or
• 50:54
  profile? That's the only thing.
• 50:55
  I'm gonna fish news. Yeah.
• 50:57
  Well The description will send it to you this news channel.
• 51:01
  So, last one that I've got for you is what is a piece of
• 51:06
  technology that you are bullish on?
• 51:09
  That would surprise people honestly.
• 51:13
  I think in general in the text-based like Hardware is kind
• 51:16
  of the next big thing. I think now software is so easy
• 51:19
  to build not use. I shouldn't say easy to build
• 51:21
  but easy and descends that a lot of the tools are very
• 51:23
  accessible. So we, we start to see a lot of
• 51:25
  the similar type of like highly My chable S3 bucket based
• 51:29
  software integrate with anything very easily.
• 51:31
  Lots of those Solutions out there and not seeing a whole lot
• 51:35
  of focus on a hardware. And I think especially working
• 51:37
  in the food space. So, hardware and packaging or
• 51:39
  the next Technologies. I'm really excited about
• 51:41
  packaging because there is a lot of opportunity on like keeping
• 51:46
  boxes, they're traveling especially for e-commerce warm
• 51:48
  or cold. So two things that haven't
• 51:50
  really been solved for yet and on the hardware side like
• 51:54
  traceability and when you think about like manufacturing for
• 51:57
  example, At lots of great traceability solutions for sort
• 52:00
  of slow manufacturing but if you think about automating like a
• 52:02
  distribution facility like an Amazon or something that's
• 52:05
  highly Dynamic, not a lot of easy Hardware Solutions out
• 52:09
  there yet. That can kind of talk to each
• 52:11
  other. That can automate what have you
• 52:14
  write? Like, sorting Pat, picking and
• 52:16
  packing with those types of operations, especially when you
• 52:18
  have a high variability of products.
• 52:21
  Let's say, like with food, right, carrot tomato, very
• 52:23
  different things. So, how that sort of like, how
• 52:26
  that Hardware sort of integrate With some things, like computer
• 52:29
  vision, for example, or reinforcement learning, and sort
• 52:32
  of having machines that can learn by themselves.
• 52:34
  I know that's a very broad topic, but I think that's what
• 52:36
  I'm most excited about and lots of opportunity in that space.
• 52:40
  That's super cool. It reminds me of a conversation
• 52:43
  we had with this guy, Gabrielle Straub, who works at ocado
• 52:48
  technology in the UK. And he talked about how like you
• 52:53
  can Envision, you know, those games that you play with the
• 52:56
  like cloth, Goes down and gets it picks.
• 52:59
  The, whatever it is stuffed animal or whatever you win.
• 53:04
  That is their warehouse. They have a ton of those, just
• 53:07
  going around and they use computer vision.
• 53:10
  They also are doing everything on the edge and it's very,
• 53:13
  they're doing it in a probabilistic way because they
• 53:16
  have to go get the apple and it's over here and then they
• 53:18
  have to go get a carrot and it's over here and they have to make
• 53:20
  sure the carrots are actually not bruised will carrots, don't
• 53:24
  bruise that much but like an avocado you want to make sure
• 53:26
  it's not Split open. And so yeah.
• 53:30
  It that is a very interesting whole like world.
• 53:36
  Not just a topic. It is a whole nother world that.
• 53:41
  Yeah, I also love to get into that, but I think that's it.
• 53:46
  Well, end it here. Any last words from you, Vishnu
• 53:51
  It's a great conversation. I think from the point of our
• 53:53
  management all the way through to talking about how data Powers
• 53:56
  hellofresh really enjoyed it. And thank you so much for
• 53:59
  joining us. Donna, thanks so much for having
• 54:00
  me guys. Great.