[00:00:04] Welcome to another episode of Tech
[00:00:06] Unhinchway Tech Gets Human. I’m your
[00:00:07] host Ravi Javeed and joining me today is
[00:00:10] Jason Cohen, VP of plant technology
[00:00:13] platform at MRO and a number one class
[00:00:15] rated leader in clinical data management
[00:00:18] serving thousands of hospitals
[00:00:20] nationwide. Jason brings over a decade
[00:00:22] in engineering, building and scaling
[00:00:24] data platforms and the teams behind them
[00:00:26] through multiple 0ero to one products.
[00:00:28] Our topic of discussion today is can AI
[00:00:31] make engineers better product thinkers?
[00:00:33] Jason, welcome to Tech Unhinged.
[00:00:35] >> Yeah, thanks for having me.
[00:00:36] >> Well, Jason, in plain terms, what do you
[00:00:38] do as VP of client technology platform
[00:00:41] at MRO and how big is the team you’re
[00:00:44] leading?
[00:00:44] >> Yeah, Clinch actually started as a
[00:00:46] clinical research recruitment platform.
[00:00:48] So, we are actually installed in many of
[00:00:51] the biggest hospitals in the country,
[00:00:53] helping them find patients for clinical
[00:00:55] trials so they can enroll more patients
[00:00:57] faster. Now that we’re a part of MRO,
[00:00:59] this large clinical data exchange
[00:01:02] company, we’re able to now leverage that
[00:01:04] data platform to provide data for other
[00:01:07] use cases such as registry, abstraction,
[00:01:10] other curation solutions. So our team is
[00:01:12] a small but mighty team of about 15
[00:01:14] people, but that team is able to service
[00:01:16] a platform that’s responsible for tens
[00:01:19] of millions of patient lives and and
[00:01:21] able to to move that data around.
[00:01:22] >> All right. So Jason, you run a clinical
[00:01:24] data platform which is one of the most
[00:01:27] regulated high stakes kind of software
[00:01:29] there is. How is the CTO’s job changing
[00:01:32] now that AI and agents are entering that
[00:01:35] world?
[00:01:35] >> Yeah, absolutely. I think before when
[00:01:38] you deliver a feature, you would have to
[00:01:40] basically deliver something that the
[00:01:42] customer was asking for in a set time
[00:01:44] frame so that way you know the business
[00:01:46] could sell that. Now it’s even more
[00:01:47] important that you’re able to put
[00:01:49] controls in place for the AI agents that
[00:01:50] are responsible for delivering these
[00:01:52] features so that can be explained on how
[00:01:54] the reasoning happened. So the way I
[00:01:56] like to think about it is any data point
[00:01:59] that we present now is something that
[00:02:00] you have to be explain in an audit a
[00:02:02] year from now is why that answer
[00:02:04] happened.
[00:02:05] >> You started your career in customer
[00:02:07] services and built yourself into an
[00:02:09] engineer and then a leader and now you
[00:02:11] were a CTO. What did starting on the
[00:02:14] customer side teach you that most
[00:02:16] engineers never get to learn?
[00:02:18] >> Yeah, I think what it taught me was a
[00:02:20] lot of what the customer wants are
[00:02:21] things that they’re not saying out loud.
[00:02:23] So, when I began my career at small
[00:02:25] adtec startup called Max Point, you
[00:02:27] know, about over a decade ago, you know,
[00:02:29] I was working with a lot of different
[00:02:30] clients, but one in particular that
[00:02:32] stood out was a frozen foods brand. They
[00:02:34] were originally targeting young families
[00:02:36] for their product, and that was not the
[00:02:38] audience that was responding according
[00:02:40] to the data. and there wasn’t a good way
[00:02:41] for us to provide them an answer at the
[00:02:43] time even though we had the data
[00:02:45] resident in our databases just just
[00:02:47] waiting to to be looked into. I
[00:02:48] eventually just went ahead and got my
[00:02:50] hands dirty rolled up my sleeves and
[00:02:52] found out that you know actually the
[00:02:53] audience that was responding the best
[00:02:55] was due to their single pack offering
[00:02:57] and that’s why it was single individual
[00:02:59] household families that were buying this
[00:03:01] not the family multiack that they
[00:03:02] thought that they would be driving sales
[00:03:04] for. So once I realized this, I built a
[00:03:05] dashboard out um that allowed them to
[00:03:07] actually slice and dice the data by skew
[00:03:09] and really really figure out what
[00:03:11] audiences were responding best to their
[00:03:13] campaign. Uh this gave them insights for
[00:03:15] you know future efforts and and kind of
[00:03:17] crossplatform marketing and all sorts of
[00:03:19] things. And this was something that
[00:03:20] turned out a lot of customers had that
[00:03:22] question in their mind that they just
[00:03:23] weren’t asking out loud. So once we
[00:03:25] built that, that was actually a reusable
[00:03:26] product that we were able to sell to
[00:03:28] multiple customers. And that’s just kind
[00:03:30] of what, you know, kind of, you know,
[00:03:31] sparked my interest in product and
[00:03:34] technology because I realized I could be
[00:03:36] much more strategic and affect change
[00:03:37] much faster if I could just get my hands
[00:03:39] on the data and tools.
[00:03:40] >> I saw that you’ve built across very
[00:03:42] different places from Red Hat, Max
[00:03:44] Point, and you know, at Clientic through
[00:03:47] the MRO acquisition. What’s something
[00:03:50] about building great technical teams
[00:03:53] that you only understood after seeing
[00:03:55] all of them?
[00:03:56] >> What I realized is it doesn’t matter how
[00:03:58] much talent you can pack or how much
[00:04:00] talent density you have on a team. What
[00:04:02] really is the biggest differentiator is
[00:04:05] how the context gets to the engineers
[00:04:08] and the ones fixing the problem. So at
[00:04:10] smaller companies, the context is super
[00:04:12] close to the engineers. The engineers
[00:04:13] are super close to the customer. They
[00:04:15] really feel their pain when something
[00:04:17] happens. uh you know there there’s a lot
[00:04:18] more 10pm emergencies trying to get
[00:04:20] something out for a client so they
[00:04:22] really understand what they’re
[00:04:22] struggling with at larger companies that
[00:04:24] have a lot more resources 24 by7 support
[00:04:27] teams and all this kind of stuff the
[00:04:28] context is much further from the
[00:04:30] engineer and I think that’s what the
[00:04:31] problem is is by the time that the work
[00:04:33] gets to the engineer we’ve played this
[00:04:34] game of telephone and now you know what
[00:04:36] we’re delivering might not be exactly
[00:04:38] what the customer was thinking but
[00:04:39] that’s no fault to the engineer or their
[00:04:40] talent or anything like that so there’s
[00:04:42] been struggles at both types of
[00:04:43] companies and you know trying to strike
[00:04:44] that balance is really where you find
[00:04:46] the high performing teams So Jason, AI
[00:04:48] has made building fast and way cheaper.
[00:04:51] And people say the hard part now isn’t
[00:04:53] how fast can we build anymore, but it’s
[00:04:56] are we building the right thing or not?
[00:04:58] So the first part of my question is as a
[00:05:00] CTO, why has product thinking become
[00:05:03] more important for engineers now and not
[00:05:06] less? I think it’s become more important
[00:05:08] because the rate at which we’re
[00:05:09] iterating allows us to be able to build
[00:05:12] a lot of these features that were an ask
[00:05:14] that was half understood last quarter.
[00:05:16] Now my engineers especially because we
[00:05:18] work on a hybrid schedule. We’re
[00:05:20] interfacing with our customerf facing
[00:05:22] counterparts on a day-to-day basis. So
[00:05:24] they’re more than happy to come over to
[00:05:25] their desk and describe the problem
[00:05:27] they’re having. And something that would
[00:05:29] be put on the backlog and you know
[00:05:31] rigorously groomed by a product manager
[00:05:33] with a the stakeholder you know months
[00:05:35] in advance of actually developing has
[00:05:37] now been compressed into a timeline
[00:05:38] where we’re finding that these
[00:05:40] conversations at the coffee machine or
[00:05:42] the water cooler actually produce the uh
[00:05:44] user story so to speak. and our our
[00:05:46] engineers are more than happy to kind of
[00:05:48] serve as that product management role
[00:05:50] and and really understand the problem,
[00:05:52] brainstorm with the stakeholder
[00:05:53] themselves and then produce that result.
[00:05:55] Just shortening that part of the
[00:05:57] development cycle has been huge and and
[00:05:59] allowed them to do that.
[00:06:01] >> So, you know, the second part to my
[00:06:02] question was that if you had to explain,
[00:06:05] you know, why product thinking matters
[00:06:07] now more than ever and if you had to
[00:06:09] explain this shift to another CTO or VP
[00:06:12] engineering in a line, how would you put
[00:06:14] it? I would frame it as the bottleneck
[00:06:16] has moved right previously the build was
[00:06:18] a long complex part and that’s that’s
[00:06:21] really what you know engineers focused
[00:06:23] on was you know how am I going to take
[00:06:24] these requirements and implement it in
[00:06:26] an iterable fashion break it into chunks
[00:06:29] and you know make it so it’s actually
[00:06:30] possible to build the thing that we’re
[00:06:32] building less so you know should we be
[00:06:35] building this why are we building this
[00:06:36] who has this problem these are the
[00:06:38] questions that the engineers need to be
[00:06:39] asking now because that build process
[00:06:41] has been compressed so that that that’s
[00:06:43] almost table stakes makes now is to be
[00:06:44] able to produce the code. The part that
[00:06:46] the engineer really needs to be able to
[00:06:47] do is translate a business problem into
[00:06:50] a technical solution. And the actual
[00:06:52] tactical way that’s done, the the
[00:06:54] individual functions and lines of code
[00:06:56] are are much less important and what’s
[00:06:58] much more important now is being able to
[00:07:00] understand why we’re solving the problem
[00:07:02] and who we’re solving it for.
[00:07:03] >> So Jason, your bio says you hire
[00:07:05] engineers who think like product
[00:07:07] managers. When you are sizing up on an
[00:07:09] engineering, what are the two or three
[00:07:12] things that tell you that they actually
[00:07:14] have that particular instinct?
[00:07:15] >> It kind of goes back to what I was just
[00:07:17] talking about. You know, if we come to
[00:07:19] an engineer with even an example problem
[00:07:21] in a in a technical interview with
[00:07:23] something they need to solve them asking
[00:07:25] product focused questions such as, you
[00:07:27] know, who raised this issue, you know,
[00:07:29] who are we solving this for? How are
[00:07:31] they working around it currently? What,
[00:07:32] you know, what kind of pain points are
[00:07:34] we trying to solve? are, you know, are
[00:07:36] are there ways that we can solve this
[00:07:38] problem and additional problems and kind
[00:07:39] of kill two birds with one stone? You
[00:07:41] know, really never being satisfied with
[00:07:43] just turning out a ticket and being able
[00:07:46] to solve a single problem, but more
[00:07:48] generically think holistically as to how
[00:07:50] are we going to build this feature that
[00:07:52] actually furthers the evolution of the
[00:07:54] product in a way that we’re able to
[00:07:56] solve similar types of problems more
[00:07:58] generically in the future instead of
[00:08:00] point solutions. So if you had to you
[00:08:02] know just bring it down to two to three
[00:08:03] things exactly do you see within an
[00:08:05] engineer that tells you that you know
[00:08:08] they are someone who can bring about
[00:08:10] that product thinking into their
[00:08:12] practice as well. Yeah, I think it’s a
[00:08:14] kind of insatiable curiosity. The ones
[00:08:17] that are asking a lot of questions and
[00:08:18] not the traditional questions that you
[00:08:20] think an engineer would ask edge cases
[00:08:23] and and how do we handle errors and
[00:08:25] failures, but more so really trying to
[00:08:27] understand the business domain and
[00:08:29] become experts in that themselves. So I
[00:08:32] would say that’s number one. Number two
[00:08:34] is customer first mentality. So always
[00:08:36] knowing that in the end it doesn’t
[00:08:38] matter how slick your tech stack is or
[00:08:41] you know how how clean the codebase is.
[00:08:44] What really matters is are we solving
[00:08:45] the the user’s problems and and do they
[00:08:48] trust us to solve their problems?
[00:08:49] >> Jason, it’s easier than ever for an
[00:08:51] engineer to get a customer answer from
[00:08:54] AI without ever talking to a customer
[00:08:56] ever. Now is that making them better or
[00:08:58] just more confident about the wrong
[00:09:00] things?
[00:09:00] >> I think it’s definitely more confident
[00:09:02] about the wrong things. You know, AI is
[00:09:04] very fluent in in trying to produce its
[00:09:07] answers. And you know, it it it sounds
[00:09:09] very confident, but it’s a lot of times
[00:09:12] very confidently wrong. I think you
[00:09:13] always have to view what the AI outputs
[00:09:16] with an adversarial lens. That’s really
[00:09:18] why that engineers are more important
[00:09:20] now than than ever because before we had
[00:09:23] deterministic inputs to produce
[00:09:25] deterministic outputs. Now both the
[00:09:27] inputs and the outputs are a bit fuzzy.
[00:09:29] So it takes an engineer with a keen eye
[00:09:31] for what could be going wrong or what
[00:09:34] could be kind of misleading from the AI
[00:09:36] and and and I think that’s the biggest
[00:09:38] challenge that we face today is is these
[00:09:39] systems are are getting extremely
[00:09:41] complex and they’re very good at being
[00:09:43] confident in the wrong direction.
[00:09:45] >> So we’re going to talk a bit about now
[00:09:46] the agents orchestration and the
[00:09:48] enterprise reality. We see that
[00:09:50] everyone’s talking about AI agents. If
[00:09:53] you were asked to give a straight answer
[00:09:55] to this one, are you actually seeing
[00:09:56] them deployed and working yet?
[00:09:58] especially somewhere as regulated as
[00:10:01] clinical data where and doing what?
[00:10:04] >> Yes, but I don’t think it’s in the way
[00:10:06] that people think of when they first
[00:10:08] hear autonomous agents. A lot of what
[00:10:10] I’ve seen, especially in regulated
[00:10:11] environments, are the agents aren’t
[00:10:13] taking autonomous actions. You know,
[00:10:15] what we have in production today is
[00:10:17] agents that assist in retrieval of data
[00:10:20] in ways that a human couldn’t possibly
[00:10:23] do. So, so think about trying to find
[00:10:25] buried deep in doctor’s notes a mention
[00:10:27] of someone having a rare disease that’s
[00:10:29] in remission. That’s not always plainly
[00:10:30] stated. It’s not coded in a structured
[00:10:32] field. It’s something that might be
[00:10:34] written across months of different notes
[00:10:36] to really get at that fact. And so, you
[00:10:39] know, the alternative is a human looking
[00:10:40] through billions of notes, which is, you
[00:10:42] know, completely impossible. The way we
[00:10:44] do it is is is have the the AI make it
[00:10:47] so it’s easily searchable to find, you
[00:10:50] know, these these insights that are, you
[00:10:51] know, a strong suggestion as to what,
[00:10:54] you know, this might have, but it still
[00:10:55] takes a human in the loop to be able to
[00:10:57] verify that that output is actually
[00:10:59] correct and take action on it. So, you
[00:11:01] know, I think in the future I could see
[00:11:03] that these agents starting to take more
[00:11:04] agency and more decision into into what
[00:11:07] they do, but right now I think it’s uh
[00:11:09] just streamlining what humans couldn’t
[00:11:12] possibly do in the compressed time frame
[00:11:14] that the agent is able to.
[00:11:15] >> Are there any other agentic use cases
[00:11:18] that you’ve observed like in your
[00:11:20] particular space other than or outside
[00:11:22] your company?
[00:11:23] >> I think everybody is is kind of rowing
[00:11:25] in the same direction as to what they
[00:11:27] want to do with these autonomous agents.
[00:11:29] So you know in the clinical data space
[00:11:31] people are trying to automate the
[00:11:33] abstraction of data. So pulling data
[00:11:35] from the form that you know a doctor
[00:11:37] writes his pros in into a form that you
[00:11:40] can put into a database that can be used
[00:11:42] for you know various applications. And I
[00:11:45] think everybody’s getting close you know
[00:11:46] they’ve automated the 70% of the easy
[00:11:49] part. It’s the 30% of what a human used
[00:11:51] to do that a clinician could explain to
[00:11:54] you why they came to that conclusion.
[00:11:56] The fuzzy steps that came to do it.
[00:11:57] every hospital does it in their own
[00:11:59] unique weird way and that’s the part
[00:12:01] that I don’t think has been solved
[00:12:03] generically yet.
[00:12:03] >> So Jason people keep saying AI
[00:12:05] orchestration like it’s pretty obvious
[00:12:08] everyone can just understand it. So you
[00:12:10] know from your technical expertise can
[00:12:12] you break down what it really means and
[00:12:14] give us one example of how it actually
[00:12:16] is working? Yeah. So, orchestration is
[00:12:18] definitely one of those things that
[00:12:19] sounds like a buzzword, but I think it
[00:12:21] actually is something that people are
[00:12:23] doing today and it’s something that’s
[00:12:24] really important to us. Uh, especially
[00:12:25] because if you’re in any type of
[00:12:27] situation that requires a latency
[00:12:29] constraint such as us where we have uh
[00:12:31] kind of near realtime requirements, uh
[00:12:33] the data has to be right and has to be
[00:12:35] prepped in advance. So, inference is
[00:12:36] happening, you know, while people are
[00:12:38] sleeping and when they wake up, they
[00:12:40] expect the answers to their questions to
[00:12:42] be there waiting for them. So, if a bad
[00:12:43] input gets in, the correction isn’t just
[00:12:45] a quick rerun. it’s hours or days of of
[00:12:48] somebody waiting on on the critical
[00:12:50] information they need to to possibly be
[00:12:52] able to service a patient. So the
[00:12:53] interesting part of engineering at this
[00:12:55] point for these agent orchestrations
[00:12:57] isn’t the model at all. It’s actually
[00:12:58] the validation between the steps and
[00:13:01] knowing how to gate something and and
[00:13:02] stop it if it’s wrong before it
[00:13:04] propagates. You know, everybody tries to
[00:13:05] talk about which model they’re using.
[00:13:07] You know, there’s all these different
[00:13:08] frontier models that boast these
[00:13:10] different benchmarks, but the expensive
[00:13:11] part is really the glue and the plumbing
[00:13:13] and the the governance and deciding at
[00:13:15] which point we can move a step forward.
[00:13:17] >> So Jason, where is the biggest gap right
[00:13:19] now between what the AI tools can do and
[00:13:22] what a regulated high stakes environment
[00:13:25] like healthcare actually needs to run
[00:13:27] this at scale.
[00:13:29] >> Yeah, and this is a big one for us and
[00:13:30] my number one answer is evidence. The
[00:13:33] tools produce an output really easily,
[00:13:35] but they don’t really produce a
[00:13:36] defensible account of how they got
[00:13:38] there. They can give you sort of a
[00:13:40] reasoning trail, something very fuzzy,
[00:13:42] but again, that can be really confident
[00:13:44] in the wrong direction. In regulated
[00:13:45] environments like ours where it it’s
[00:13:47] hipporegulated, you have to make sure
[00:13:49] that the reasoning trail is a
[00:13:51] deliverable, auditable trail. And it’s
[00:13:54] definitely not one of these nice to
[00:13:56] haves. It’s a requirement that lies
[00:13:58] underneath any feature request that
[00:14:00] involves AI, whether the person
[00:14:02] requesting that explicitly states that
[00:14:04] or not.
[00:14:04] >> Engineers first used to get promoted for
[00:14:07] building complex things. But if AI takes
[00:14:09] a chunk of that thing away, what should
[00:14:12] they get promoted for now?
[00:14:14] >> Yeah. Yeah. And I I think I think you’re
[00:14:16] right. Like, you know, it used to be
[00:14:18] based off of the complexity or the size
[00:14:20] of the things they built. You know, I
[00:14:23] even remember early in my career, people
[00:14:24] were still counting lines of code, but
[00:14:26] I’ve always thought those were weird
[00:14:27] proxies. You know, they worked at the
[00:14:29] time because the complexity of writing
[00:14:31] code and building something was
[00:14:32] expensive. So surviving kind of through
[00:14:35] that era kind of proved something that
[00:14:38] you could deliver. But now I think it’s
[00:14:39] a lot about judgment. Is that a engineer
[00:14:41] able to use kind of that product
[00:14:43] judgment? Are they able to judge in a
[00:14:46] feature request what really is at the
[00:14:48] root of the problem and how we should
[00:14:49] solve it? Because before the solution
[00:14:51] that you know they would produce for
[00:14:53] something an MVP was kind of driven by
[00:14:55] the the expensiveness of the of actually
[00:14:58] implementing the solution whereas now
[00:15:00] that’s the part that’s been compressed
[00:15:01] and that’s the part that’s super cheap.
[00:15:03] So I do think engineers are are now
[00:15:05] promoted for their ability to translate
[00:15:07] customer requirements into a working
[00:15:09] product. So kind of bringing both the
[00:15:12] worlds of product and engineering
[00:15:13] together. you know product managers
[00:15:15] closer to the metal engineers closer to
[00:15:17] the customer kind of in this uh you know
[00:15:20] hybrid role so to speak I know a lot of
[00:15:22] companies talk about forward deployed
[00:15:24] engineers you know my teams coming from
[00:15:26] startups have generally always been very
[00:15:28] very close to the customer and so I
[00:15:29] think in today’s age now kind of a
[00:15:31] requirement even in larger companies so
[00:15:33] I think that’s what they get promoted
[00:15:34] for is is for being able to use their
[00:15:37] judgment to produce results
[00:15:38] >> so Jason as an engineer what are your
[00:15:41] top three go-to AI tools right now Yeah.
[00:15:43] So that was actually one of the biggest
[00:15:45] things for me is not trying at this
[00:15:47] point not getting locked into a single
[00:15:49] tool. We’re kind of a multi-tool shop
[00:15:51] here. Use each tool for for its best
[00:15:53] purpose. So we use Open AAI and Codeex
[00:15:56] is kind of our everyday workhorse. You
[00:15:57] know, it’s really good at getting things
[00:15:59] done and not over complicating. But if
[00:16:00] we have some massive abstract problem
[00:16:03] that we need to solve, we’re really
[00:16:05] finding Claude code to be, you know,
[00:16:07] really good at at planning and
[00:16:09] ideulating and and that kind of a thing.
[00:16:11] And so that’s more of our, you know,
[00:16:13] specialty tool that we bring out in
[00:16:15] certain situations. As as far as a
[00:16:17] third, that’s to be seen. We’re
[00:16:19] experimenting with, you know, with with
[00:16:20] a lot of other different tools and
[00:16:22] models and providers. And and we’re
[00:16:24] noticing kind of a compression of, you
[00:16:26] know, what you need a a Frontier model
[00:16:29] for versus what, you know, something
[00:16:31] like a Gemini Flash Light can do for
[00:16:34] you, right? It’s all about context
[00:16:36] engineering and and and how you’re
[00:16:38] framing the problem. and and we’re
[00:16:39] finding that a lot of these I guess kind
[00:16:42] of lower tier models can produce better
[00:16:44] results in the hands of a you know more
[00:16:46] experienced engineer than these front
[00:16:49] tier models could with with someone
[00:16:50] who’s kind of trying to oneshot prompt
[00:16:52] it. So Jason now before you know we come
[00:16:54] to the conclusion um and our last
[00:16:56] question what do you think can AI make
[00:16:59] engineers better product thinkers is
[00:17:01] that a yes or a no and why
[00:17:03] >> I think it’s an absolutely yes uh I
[00:17:06] think that they have to based off the
[00:17:08] the way the world is changing right like
[00:17:10] no longer can engineers kind of be
[00:17:12] sitting in the back room and you know
[00:17:14] headphones on heads down just cranking
[00:17:16] out code because that’s the part that is
[00:17:18] no longer expensive and no longer takes
[00:17:20] a lot of time I think they have to be
[00:17:23] able to think like a product thinker to
[00:17:25] be able to accurately describe to the
[00:17:27] agent what needs to be built. Gives them
[00:17:29] the sense of what they need to be
[00:17:30] testing for and checking for. And I
[00:17:33] think overall I, you know, we’re seeing
[00:17:35] again that convergence to the middle of
[00:17:37] of of the roles of product and
[00:17:38] engineering. We’re we’re no longer our
[00:17:41] engineers, you know, specific code
[00:17:43] specialists. Um, you know, even product
[00:17:45] managers now can write some simple code
[00:17:47] to build prototypes. So the the coding
[00:17:49] is no longer the barrier. barrier is who
[00:17:51] can grab a problem and harness it and
[00:17:53] actually produce a result that you know
[00:17:55] makes the customer happy.
[00:17:56] >> So Jason, you’ve led teams through
[00:17:58] multiple 0 to1 bills and climbed an
[00:18:00] unusual road to being a CTO. What’s the
[00:18:03] hardest thing you’ve had to unlearn as a
[00:18:06] leader and what advice would you give
[00:18:08] our listeners today? Yeah, I think you
[00:18:11] know this actually took me much longer
[00:18:12] to learn than than it probably should
[00:18:14] have. But coming from a non-traditional
[00:18:16] background in terms of engineering and
[00:18:18] not originally having that computer
[00:18:20] science credential, I always felt like I
[00:18:22] had to be the technically strongest in
[00:18:24] the room even as a leader. You know, it
[00:18:25] it turns out that’s not the best way to
[00:18:27] lead. You know, empowering your team to
[00:18:30] be able to make autonomous decisions on
[00:18:32] their own is a sign that you’ve built a
[00:18:34] good team and a good structure and a
[00:18:36] good process. So, you know, I had to
[00:18:38] unlearn that. I think even leaders that
[00:18:40] come from a technical background, it’s
[00:18:42] something that’s hard to let go of
[00:18:44] because they’ve often come up through
[00:18:45] the ranks and, you know, we’re that tech
[00:18:47] leader, we’re that, you know, or expert
[00:18:49] on their team. But at some point, you
[00:18:51] you’ve got to let that go. And you learn
[00:18:53] that leadership is just building a
[00:18:55] culture that can foster building
[00:18:57] engineers that are technical champions
[00:18:59] and you know kind of psychological
[00:19:01] safety to be able to ask these questions
[00:19:03] that we now have to ask to the customers
[00:19:06] directly or you know get get more into
[00:19:08] the weeds on on things that were
[00:19:10] traditional product conversations that
[00:19:12] they were having separate from the
[00:19:13] engineering team.
[00:19:14] >> Yeah. I know that that was great advice
[00:19:16] and you know some things are best to be
[00:19:18] unlearned you know as long as we want to
[00:19:20] move forward with our lives and in the
[00:19:22] best possible ways being leaders. So
[00:19:24] Jason thank you so much for your
[00:19:26] insightfulness uh for our listeners
[00:19:28] today. It’s been a pleasure having you
[00:19:30] on Tech Unhinged.
[00:19:31] >> Yeah thanks for having me. I really
[00:19:32] appreciate it.