The Tech Trek

• Factory operating systems and the AI hardware crunch

  Karan Talati, cofounder and CEO at First Resonance, joins me to unpack what modern manufacturing really looks like inside factories that build rockets, drones, reactors, and other complex hardware. We dig into why only a small slice of factories run on real systems today, what a true factory operating system unlocks, and how that connects directly to national security and the AI boom.If you care about where all of this new compute, energy, and defense hardware will actually come from, this conversation gives you a clear view of the stack, the gaps, and the opportunity. Key takeaways• Only a small fraction of factories in the United States use a manufacturing execution system, which leaves a huge gap between legacy on prem tools, paper processes, and generic workflow apps that were never built for hardware work• Cloud infrastructure and open interfaces now make it possible to deploy a purpose built factory operating system at a cost and speed that works for both fast moving startups and long standing suppliers• Reindustrialization does not mean bringing every product back onshore, it means being deliberate about the layers of manufacturing that matter most for national security, chips, optics, and other high value components• The real foundation for modern manufacturing is talent, there is a major chance to re skill people into highly technical, well paid roles in aerospace, semiconductors, energy, and more• AI and agent style workflows will sit across design, manufacturing, and field operations so that hardware teams can close feedback loops, shorten timelines, and make better decisions with the data they already generateTimestamped highlights[00:40] Karan explains what First Resonance does and why he calls it a factory operating system for complex industries like aerospace, defense, energy, and autonomy[01:55] How we ended up with only about fifteen percent of factories running on an MES, and why most hardware work still lives on paper, spreadsheets, and ad hoc tools[06:49] A clear walkthrough of how offshoring looked like a rational path for decades, and why it created hidden risk across chips, optics, and other critical components[11:46] Which parts of manufacturing should come back onshore, why you do not want everything local, and how workforce strategy fits into the new industrial map[16:35] What a horizontal stack across design, factory systems, test, and field data can look like, and how AI agents can keep teams in sync across that stack[23:02] The real timelines of hardware in the age of AI, why software is speeding up physical development, and why examples like SpaceX and TSMC matter for the next decadeA line that stayed with me“Hardware and software are not separate worlds, they are one system that is now converging faster than most people realize.”Practical moves for tech leaders• Map your current manufacturing and hardware workflows, even if you are at a software first company, find the paper, spreadsheets, and disconnected tools that support anything physical you ship• Look for one or two places where a factory operating system or modern MES could remove handoffs, for example design changes that take weeks to reach the line or test data that never feeds back into engineering• Treat manufacturing careers as part of your talent strategy, help your teams see these roles as high skill and high impact, not as a side trackCall to actionIf this episode gave you a clearer view of how hardware, AI, and national security tie together, share it with one other person who should be thinking about the factory side of their roadmap. Follow and subscribe to The Tech Trek so you never miss deep dives like this, and connect with me on LinkedIn if you want more conversations at the edge of data, engineering, and real world impact.

  --------  

  28:04

  --------

  28:04
• Inside the Business of Modern Waste Management

  Michael Marmo, founder and chief executive of CurbWaste, joins The Tech Trek to share how he went from catching fastballs in Europe to building software that runs the daily work of waste haulers. We walk through the very human side of leaving a sports identity, starting at the bottom in a family waste business, and finally asking a simple question about founding a company. Why not meIf you are sitting inside an industry and quietly seeing the gaps that no product seems to solve, this conversation is a playbook in how to turn that insider view into a real business, even if you do not come from a traditional tech background.Key takeaways• Identity can change, but the work habits that made you good at sports or any craft can transfer directly into building a company, especially persistence, dealing with failure, and showing up every day• You do not have to love a specific activity forever, you can follow the deeper thread underneath it, like merit, teamwork, and visible impact, and find those same traits in a very different industry• Deep time inside an industry lets you see painful, repeatable problems, and that is often a better seed for a product business than starting with a clever idea and pivoting until something sticks• A clear why for the product and a clear why you are the person to build it are not nice to have, they are what convince customers, hires, and investors to follow you when things get hard• Great founders do not pretend to be good at everything, they are honest about what they do not know, learn just enough to make good calls in product, engineering, and go to market, and then surround themselves with people who fill the gapsTimestamped highlights00:32 Michael explains what CurbWaste does and how it runs a hauler business from first customer contact through billing01:21 From college baseball and pro teams in Europe to the first job in media and tech sales, and the identity shock that came with that change06:27 What it really felt like when the game ended, why mens leagues did not scratch the itch, and how that led to a quiet reset in the working world09:11 Starting at the bottom in a family recycling center, discovering a love for the waste industry, and why it felt like a merit based team environment15:24 Walking the floor at Waste Expo, not finding the software he needed, deciding to fund and build his own tools, and seeing other haulers facing the same problems19:40 The moment hearing the Yelp founder speak turned into a personal question, why not me, and how that idea of trying anyway shapes the way he thinks about founding todayA line that stayed with me“At the end of the day he tried. He had an idea and he acted on it and pursued it. That really resonated. I was like, why not me”Practical notes for future founders• Before you write any code or quit your job, write down why this problem matters, why it matters now, and why you are willing to keep going when it stops being fun• If your first answer to why is only about money, keep digging until you find something that still feels true on a hard day, because you will have a lot of those• Use your current role as a live lab, list the moments that feel broken, expensive, or slow, and ask which of those could actually support a business if you solved them well• Be direct with yourself about weak spots, whether that is product, tech, or selling, then build a basic understanding and lean on people who are strong where you are notCall to actionIf you enjoy stories that get inside how real founders make the leap from operator to builder, follow The Tech Trek in your favorite podcast app and share this episode with someone who is quietly thinking about starting something of their own.

  --------  

  25:06

  --------

  25:06
• How data teams are rebuilding insurance from the inside

  Jason Ash, Chief of Data at Symetra, joins the show to unpack how a mid sized insurer is rebuilding its data stack and culture so business and technology actually pull in the same direction. He shares how his team brings actuaries, product leaders, and engineers into one data platform, and why opening that platform to non technical contributors has been a turning point. If you work in a regulated industry and are trying to move faster with data, this conversation gives you a very practical view of what it takes.Key takeaways• Business and tech only work when they share context and trustJason has sat in both seats, first as an actuary and now as a data and engineering leader. That dual background helps him translate between risk, regulation, and modern data practices, and it shapes how he frames projects around shared business outcomes rather than tools.• Put data leaders inside business line leadership, not on the outsideSeveral of Jason’s managers sit on the leadership teams for Symetra’s life, retirement, and group benefits divisions. They hear priorities and constraints at the same time as product and distribution leaders, which lets them frame data as a value add for new products instead of a back office cost.• Treat the warehouse as a shared product and measure contributors, not just tablesSymetra’s dbt based warehouse started with about five contributors. Over three years they grew that to more than sixty, and half of those people sit outside the core data team. Business users learn to contribute SQL, documentation, and domain knowledge directly into the repo, which spreads ownership and reduces bottlenecks.• Shift stakeholders away from big bang launches to steady deliveryJason pushes his teams to think like software engineers. Rather than promising a perfect data product on a single date, they deliver an early slice of data, have partners use it right away, collect feedback, and improve every month. That builds trust and avoids the usual disappointment that comes with one big release.• Use maturity as a guide for where to investEarly on, his group picked a few strong champions who were willing to accept slower delivery in exchange for building real infrastructure. Now that the platform and practices are in place, the focus is on scale, reuse, and getting more people to build on the same foundation, including as AI capabilities start to reshape the work.Timestamped highlights00:53 Jason explains what Symetra actually does and how their product mix makes data work more complex than the company size might suggest02:19 From actuary to Chief of Data, and what sitting on both sides of the fence taught him about business and technology expectations08:08 Why mixing data engineers, data scientists, actuaries, and analysts on the same problems leads to stronger solutions than any single discipline alone13:44 How embedding data leaders into each business division’s leadership group changed when and how data enters product discussions16:38 The dbt story at Symetra, and how more than sixty people across the company now contribute directly to the shared data warehouse26:22 Moving away from big bang data launches and setting expectations around early value, continuous feedback, and ongoing quality improvements32:06 The tension between safety and speed as AI advances, and what Jason worries about most for established insurers that move too slowlyPractical moves you can steal• Put data leaders on business line leadership teams so they hear priorities and constraints in real time, not after the roadmap is set• Track how many unique people contribute to your data warehouse and make that a visible success metric across the companyStay connectedIf this episode helped you think differently about data leadership in regulated industries, share it with a colleague who owns product, data, or actuarial work.

  --------  

  38:14

  --------

  38:14
• Data Culture That Actually Delivers With AI

  Chris Morgan, VP of Data Science at Lincoln Financial Group, joins me to unpack what a real data culture looks like inside a complex, highly regulated business that has policies on the books for decades. We talk about how to turn Gen AI buzz into real value, why governance and quality suddenly matter to everyone, and how to tackle data technical debt without stalling delivery.Chris shares concrete ways he finds champions in the business, balances centralized and federated models, and keeps stakeholders excited about the future while he quietly fixes the messy data foundation underneath it all.Key takeawaysData culture is less about dashboards and more about curiosity, repeatable processes, and raising the analytical watermark across the company, not just in the data team.The teams that will win with Gen AI are the ones that can safely connect proprietary data to these models, which demands strong governance, clear definitions, and shared standards.A blended model works best for scaling data work, where a central function sets guardrails and standards while domain teams stay close to the business and own local decisions.Paying down technical debt works when it is framed in business terms, tied to revenue and risk, and treated as a regular slice of capacity instead of a one time side project.Education is now part of the job for data leaders, from internal road shows on Gen AI to simple stories that explain why foundational data work matters before you can ship shiny tools.Timestamped highlights00:04 Setting the stage Chris explains his role at Lincoln Financial and how data science supports life and annuity products that can live for decades.03:33 The Cobb salad story A simple grocery store analogy that makes data standards and shared definitions instantly clear to non technical stakeholders.06:06 Finding the right champions Why Chris prefers curious partners who will invest time with the data team over senior leaders who just want results without changing behavior.08:33 Governance as Gen AI fuel How regulatory pressure and the need to trust what goes into models are pushing data governance and quality into the spotlight.11:11 A practical way to attack data technical debt How Chris decides what to fix first, and why he tries to reserve a steady slice of team time for cleanup so progress is visible and sustainable.17:44 Managing Gen AI expectations From road shows to constant communication, Chris shares how he keeps enthusiasm high while also being honest about the timeline and effort.One line that sums it up“These generative models are going to become a commodity and what will separate companies is who can take the most advantage of their proprietary data.”Practical playbookStart small with data culture by picking one engaged business partner, one problem, and one outcome you can measure clearly.Reserve a consistent portion of team capacity for technical debt, even if it is only a small percentage at first, and make the tradeoffs visible.Use stories, analogies, and simple rules of the road so stakeholders can understand how data systems work without becoming experts in the tech.Call to actionIf this conversation helped you think differently about data culture and Gen AI inside your company, follow the show and leave a rating so more engineering and data leaders can find it. To keep the discussion going, connect with me on LinkedIn and share how your team is tackling data culture and technical debt right now.

  --------  

  28:00

  --------

  28:00
• How AI Role Play Levels Up Public Speaking Interviews and Tough Conversations

  Varun Puri, CEO and cofounder of Yoodli, joins the show to talk about using AI role play to transform how people practice for high stakes conversations, from sales calls to job interviews to tough manager chats. He breaks down how Yoodli went from a consumer public speaking tool to a serious enterprise platform used by teams at Google, Snowflake, Databricks, and more, all while staying anchored in one mission, helping humans communicate with confidence. We dig into product led growth, honest feedback loops, and why real human communication will matter even more as AI makes information instant.Key takeaways• Why Yoodli started with public speaking anxiety and grew into an AI role play simulator for any important conversation, not just conference talks or pitch decks• How watching real user behavior inside companies like Google pulled the team into enterprise without abandoning their consumer product• A simple approach to product feedback, talk to end users constantly, then prioritize changes by business impact, renewal risk, and how many people benefit• What it really takes to move from consumer to enterprise, new roles, new processes, and a very different mindset around reliability, security, and expectations• Why Varun draws clear ethical lines, using AI to coach and prepare people, not to replace human judgment in hiring, promotion, or high trust decisionsTimestamped highlights[00:35] What Yoodli actually does today, from solo practice to training sales and go to market teams inside large enterprises[01:43] The original vision, helping people who are scared of public speaking, and the insight that interviews, sales calls, and manager talks are all just role plays[03:37] How the team listens to end users, the channels they rely on, and why the consumer product is still their testing ground for new ideas and experiments[05:20] Following users into the enterprise, why it was an addition and not a full pivot, and how product led growth inside companies like Google works in practice[07:42] The early shock of selling to enterprises, learning about new roles, SLAs, InfoSec, and bringing in leaders from Tableau and Salesforce to build a real B2B engine[11:10] Two paths for AI in sales, tools that try to replace humans versus tools that make humans better, and why Varun has drawn a hard line on what Yoodli will not do[15:26] A future where information is commoditized and instant, and why communication and presence become the real edge for top performers in that world[20:48] Designing for trust and adoption, how Yoodli keeps practice private by default, when data is shared, and why control has to sit with the end userA line worth saving“In a world where AI makes everyone smarter and faster, the thing that will be at the biggest premium is how you communicate as a human with other humans.”Practical ideas you can use• Keep a consumer like surface in your product so you can experiment faster than your enterprise roadmap would ever allow• Treat feedback from large customers like a queue you rank by renewal risk, strategic value, and number of users helped, not as a list you must clear• Look for product led growth signals inside your user base, if thousands of people in one company are using you, someone there probably wants a team level solution• Draw explicit boundaries for your AI product, write down what you will not automate, so you can build trust with users and buyers over the long termCall to actionIf you care about the future of sales, interviewing, and communication in an AI rich world, this conversation is worth a listen. Follow the show, leave a quick rating, and share this episode with a founder, product leader, or sales leader who is thinking about AI in their workflow. And if you want feedback on your own speaking, check out what Varun and his team are building at Yoodli.

  --------  

  24:09

  --------

  24:09

More Technology podcasts

Trending Technology podcasts

About The Tech Trek

The Tech Trek: Podcasts in Family