The Tech Trek

Vipin Kumar, Head of CUSO IB Data Strategy and Analytics at Deutsche Bank, joins me to unpack one of the toughest problems in financial services: managing data quality in a highly regulated industry. From the outside, it might look like a box-checking exercise. In reality, it’s a complex mix of legacy systems, global frameworks, regulatory controls, and the constant push to balance defensive compliance with offensive business value. Vipin makes it real with examples that connect directly to how we all experience data in daily life.

Key Takeaways

Data quality isn’t just about accuracy—timeliness, completeness, and consistency all matter, especially when billions are on the line.

Regulations push banks into “defensive” strategies, but there’s growing opportunity to apply “offensive” strategies that use data for prediction, analytics, and competitive edge.

Measuring effectiveness requires agreement between data producers and consumers, with preventive and detective controls working together.

AI and machine learning are starting to automate checks, spot patterns, and even strengthen anti-money laundering defenses.

Timestamped Highlights

00:45 What data quality means in a regulated industry

03:15 The challenges of managing fragmented legacy systems

06:40 How producers and consumers measure effectiveness of frameworks

09:30 The pizza delivery analogy for making sense of data quality

14:20 Why accuracy is harder than timeliness or completeness

16:50 The role of AI and machine learning in improving governance

19:20 Shifting from defensive compliance to offensive strategy in banking

22:40 Regulators testing AI-driven approaches to anti-money laundering

Memorable Quote

“Producer has preventive controls. Consumer has detective controls. True data quality happens only when both align 100%.” — Vipin Kumar

Call to Action

If you enjoyed this conversation, share it with a colleague who thinks about data quality or governance. Don’t forget to follow the show on Apple Podcasts or Spotify so you never miss an episode.

Marty Ringlein, co-founder and CEO of Agree.com, joins Amir to unpack why history always repeats itself in technology and what that means for the AI era. From the telephone to the automobile to ChatGPT, the biggest shifts have rarely been things people asked for—they were inventions that reshaped behavior once adopted. Marty explains why skepticism always comes first, how fear fuels resistance, and why optimism is usually rewarded. He also shares how Agree.com is rethinking contracts and payments by automating the painful parts of sales workflows.

Key Takeaways

The most transformative inventions weren’t requested—they emerged through evolution and network effects.

Human resistance to new tech often comes from energy costs of relearning, not the tech itself.

AI isn’t eliminating jobs—it’s freeing people from low-value work so they can focus on bigger challenges.

Every wave of disruption (printing press, cars, internet, mobile, AI) begins with fear, then proves to be a net positive.

Timestamped Highlights

00:51 — Why Agree.com calls itself “a better DocuSign” and how it integrates signatures, invoicing, and payments

02:06 — The history of inventions nobody asked for and why they stuck

05:41 — Human pessimism vs optimism when confronting new technologies

09:05 — Why fears around AI echo the same debates once had about books, cars, and the cloud

13:38 — How automation frees salespeople and engineers to focus on higher-value work

18:51 — Are there technologies that have been net negative for society? Marty’s take

23:21 — Why every generation thinks “this time it’s different”

Memorable Quote

“The biggest things that will change our lives are the ones we don’t even know to ask for yet.” — Marty Ringlein

Call to Action

If you enjoyed this episode, share it with a colleague who’s navigating the AI conversation. Follow The Tech Trek for more conversations that cut through the noise on tech, leadership, and the future of work.

Simon Lam, VP of Engineering at M1, joins the show to unpack one of the trickiest topics in tech careers: how engineers can build influence without a formal leadership title. Too often, influence is mistaken for charisma or public speaking—but Simon explains why it’s really about consistent impact, trust, and understanding how change happens inside teams. If you’re an IC who feels stuck at the “senior wall” or a manager wondering how to better evaluate career growth, this conversation delivers clarity and actionable insight.

Key Takeaways

• Influence isn’t charisma—it’s the result of consistent impact and trust over time

• Engineers can build influence at any stage, from junior to staff, by solving problems and being reliable

• Career progression should tie back to impact, not just who speaks the loudest in the room

• Change management offers a practical lens for understanding influence in technical settings

• Dual career tracks mean engineers don’t need to move into people management to keep advancing

Timestamped Highlights

01:39 Why influence is often misunderstood in engineering careers

05:12 Influence vs charisma—and why you don’t need to be an extrovert

08:47 The virtuous cycle of impact leading to influence

13:20 Are companies biased toward rewarding outspoken engineers?

17:21 Practical ways ICs can start building impact today

22:48 Why you don’t need to manage people to have a leadership career

A line worth remembering

“Consistent impact is how you build influence.” — Simon Lam

Call to Action

If this episode sparked new ways to think about your own career, share it with a teammate who’s navigating the same questions. Follow the show for more conversations with leaders shaping the future of engineering.

CJ King, CTO at Torc Robotics, joins the show to talk about the future of autonomous trucking at scale. Instead of asking “can we build one self-driving truck?” Torc is asking, “how do we safely put 10,000 on the road?” From supply chain transformation to regulatory hurdles, CJ breaks down what it really takes to bring production-ready autonomous semis into the market and why the ripple effects will reach far beyond trucking.

Key Takeaways

• Scaling autonomous vehicles isn’t about prototypes—it’s about building production-ready systems from the ground up.

• Trucks face unique technical challenges, from 1,000-meter perception needs to fully redundant systems that can’t rely on cloud compute.

• Removing driver limitations could extend operations from 8 hours a day to 20, unlocking major gains in supply chain efficiency.

• Regulatory collaboration is critical—success depends on alignment with federal and state agencies, law enforcement, and logistics partners.

• Adoption will come in step-functions: once proven safe and reliable, logistics companies are ready to adopt at scale.

Timestamped Highlights

00:45 – Torc’s focus on hub-to-hub autonomous trucking

02:03 – Why scaling to thousands of trucks matters more than building one prototype

06:48 – The unique technical problems of trucks vs. passenger cars

09:25 – How extended operating hours reshape logistics and supply chains

14:17 – Working with regulators and law enforcement to ensure safety and compliance

17:42 – AV3.0, synthetic data, and billions of miles of training for safer systems

22:31 – Building public trust and societal acceptance of autonomous trucking

25:21 – Why large-scale adoption will happen in step functions, not trickles

A Line That Stuck With Us

“Our bare minimum is to drive as good as a human—our mission is to be safer than one.” – CJ King

Call to Action

If you enjoyed this episode, share it with someone who cares about the future of tech and logistics. Make sure to follow the show so you never miss conversations that dig into how technology is reshaping our world.

Joseph Krause, co-founder and CEO of Radical AI, joins the show to break down how scientific discovery is being reinvented. From the limitations of the traditional trial-and-error model to the rise of AI-driven self-driving labs, Joseph explains how science is moving from slow, serial processes to a parallel model that unlocks breakthroughs at scale. He also dives into the economics of materials, why big companies can’t pivot fast enough, and how the role of scientists is being transformed.

Key Takeaways

The old model of science is serial: slow, linear, and limited by human capacity to read, experiment, and analyze.

Negative results—failed experiments—are the true fuel for breakthroughs, but they’re rarely captured or shared.

Self-driving labs powered by AI create a “materials flywheel,” running 30,000+ experiments a year and learning continuously.

Big corporations are trapped by the innovator’s dilemma and talent challenges, leaving space for startups to lead.

Scientists in the future will focus less on repetitive lab work and more on shaping hypotheses and applying intuition at scale.

Timestamped Highlights

02:00 How science traditionally works and why it’s so slow

05:50 Why mistakes and negative results matter more than we admit

09:40 The fragmentation of research and why labs don’t share data

17:15 Inside a self-driving lab and how AI accelerates discovery

23:40 Why big material companies can’t innovate like startups

35:40 The new role of scientists in an AI-powered discovery world

Memorable Line

“You don’t get a PhD to learn to pipette—you get it to think about how and why the world will change.”

Call to Action

If you enjoyed this conversation, share it with a colleague who geeks out on science and technology. Follow the show on Apple Podcasts or Spotify so you don’t miss future episodes exploring where tech is headed next.

John Fiedler, SVP of Engineering and CISO at Ironclad, joins the show to unpack the real challenges of technology leadership. From managing nonstop context switching to measuring success when you’re no longer shipping code, John shares hard-earned lessons on how leaders can protect their time, set priorities, and thrive in the chaos. Whether you’re moving from IC to manager or scaling as an executive, this conversation offers a candid look at what it truly takes to lead.

Key Takeaways

• Success in leadership isn’t about features shipped—it’s about execution, people, and culture.

• Context switching is constant, but leaders can design their calendars to minimize the chaos.

• Organizational size reshapes the challenge: startups reward speed, enterprises demand process.

• Protecting your time isn’t optional—leaders who don’t own their calendars quickly burn out.

• The leap from IC to manager requires starting fresh and mastering a new craft.

Timestamped Highlights

02:13 The hidden tax of context switching

06:53 How John measures success as a leader without code

10:45 What really slows executives down inside organizations

15:51 How John protects his calendar and finds focus time

24:47 The lessons every first-time manager needs to hear

A Line That Sticks

“If you don’t control your calendar, your calendar will control you.”

Call to Action

If this episode resonated, share it with a fellow leader navigating the chaos. Subscribe to The Tech Trek on Apple Podcasts and Spotify for more candid conversations about scaling, leadership, and the future of technology.

Alex Salazar, co-founder and CEO of Arcade.dev, joins the show to unpack the realities of building enterprise agents. Conceptually simple but technically hard, agents are reshaping how companies think about workflow automation, security, and human-in-the-loop design. Alex shares why moving from proof-of-concept to production is so challenging, what playbooks actually work, and how enterprises can avoid wasting time and money as this technology accelerates faster than any previous wave.

Key Takeaways

Enterprise agents aren’t chatbots—they’re workflow systems that can take secure, authorized actions.

The real challenge isn’t just building demos but getting to production-grade consistency and accuracy.

Mid-market companies face the steepest climb: limited budgets, limited ML expertise, but the same competitive pressure.

Success starts with finding low-risk, high-impact opportunities and narrowing scope as much as possible.

Authorization is the biggest blocker today; delegated OAuth models are key to unlocking real agent functionality.

Timestamped Highlights

02:02 — Why agents are “just advanced workflow software” but harder to trust than traditional apps

04:53 — The gap between glorified chatbots and real enterprise agents that take action

09:58 — From cloud mistrust to wire transfers: how comfort with automation evolves

14:00 — Chaos at every tier: startups, enterprises, and why the mid-market struggles most

26:21 — The playbook: how to pick use cases, narrow scope, and carry pilots all the way to prod

34:38 — Breaking down agent authorization and why most RAG systems fail in practice

42:09 — Adoption at double speed: what makes this AI wave different from internet and cloud

A Thought That Stuck

“An agent isn’t an agent until it can take action. If all it does is talk, it’s just a chatbot.” — Alex Salazar

Call to Action

If this episode gave you a clearer lens on enterprise agents, share it with a colleague who needs to hear it. And don’t miss future conversations—follow The Tech Trek on Apple Podcasts, Spotify, or wherever you listen.

Russ d’Sa, founder and CEO of LiveKit, joins the show to unpack the rise of voice AI and what it means for how we interact with technology. From the shift away from static decision trees to dynamic, LLM-powered systems, Russ explains why voice is emerging as one of the most natural interfaces for humans—and one of the most disruptive opportunities for builders. This episode goes beyond surface-level hype to explore real-world use cases, infrastructure shifts, and what’s coming next as voice moves from novelty to mainstream.

Key Takeaways

• Voice AI has moved far beyond Siri and Alexa—LLMs enable open-ended, natural conversations without rigid decision trees.

• Two main categories are emerging: open-ended voice experiences (like tutoring and therapy apps) and goal-oriented workflows (like healthcare intake, finance, and customer support).

• The biggest barrier isn’t just technology, but adoption behavior—older generations default to typing and screens, while younger users and voice-first cultures are accelerating change.

• Infrastructure for voice and video AI requires a fundamental shift from stateless web servers to stateful, long-lived conversational systems.

• The hardest technical challenge ahead: mastering conversational turn-taking so AI can interact as naturally as a human.

Timestamped Highlights

01:06 How LiveKit is giving applications the ability to see, hear, and speak

04:18 The two main categories of voice AI use cases emerging right now

09:53 Why adoption of voice AI depends as much on behavior as on technology

14:20 Imagining a 24/7 voice-driven AI that replaces screens and UIs

20:30 Why the internet’s original infrastructure wasn’t built for voice and video AI

25:39 The challenge of memory, authentication, and group dynamics in AI conversations

A line worth remembering

“If you have a computer that perfectly understands when to speak, when to listen, and adds value in the right moments—why would you ever use anything else?”

Call to Action

If you enjoyed this conversation, share it with a colleague who’s curious about where AI is headed. Subscribe on Apple Podcasts or Spotify so you don’t miss future episodes diving into the technologies shaping the next decade.

Sumit Arora, VP of Advanced Technology at Ascend Learning, joins the show to unpack the real challenges of turning AI prototypes into production-ready systems. From managing non-deterministic outputs to rethinking the relationship between engineering and product, Sumit shares hard-earned lessons on what it actually takes to build AI that works at scale. If you’re navigating how to move beyond experiments and deliver AI products that stick, this episode will give you a clear look at the path forward.

Key Takeaways

• Scaling AI is not about building smarter prototypes—it’s about mastering distributed systems, security, and availability.

• The best AI teams combine deep systems engineering with practical product sense.

• Traditional software requirements processes won’t work for AI. Co-creation between product and engineering is essential.

• Innovation pods—small, cross-functional teams—can accelerate experimentation without killing momentum.

• Success at scale comes from modular, reusable AI systems that can plug into multiple contexts.

Timestamped Highlights

02:14 — Why building a working AI demo is easy, but scaling it into a reliable product is hard

04:49 — Lessons from the big data revolution and how AI is moving even faster

08:41 — The skill sets AI teams really need and why distributed systems expertise trumps pure ML

13:13 — Designing user experiences for AI and why response times redefine UX expectations

17:00 — The evolving relationship between product and engineering in the AI era

23:10 — How innovation pods help organizations experiment without stalling production teams

26:47 — Why modular, self-contained AI systems are the key to scaling across an enterprise

A Line That Stuck

“You can’t requirement doc your way to AI success. Product and engineering have to co-create and move fast.”

Call to Action

If you found this conversation useful, share it with a colleague, subscribe to the show, and leave a quick rating—it helps us bring more tech leaders and practitioners to the table.

Sean McCarthy, co-founder and CEO of BackOps, shares how a career in sales prepared him to build an AI-driven logistics company from the ground up. In this episode, Sean reveals how observing real-world pain points at Amazon inspired BackOps’ mission and why coming from a non-technical background can actually be a founder’s advantage. This is a conversation about scaling, selling, and leading with insight — perfect for anyone thinking about making the leap from operator to founder.

Key Takeaways

• Why non-technical founders are uniquely positioned to solve operational problems with AI

• The mindset shift required to go from running sales to running an entire company

• How to validate an idea before leaving a stable, well-paying job

• What it really takes to hand off sales when it’s been your superpower

• Pricing insights that help ensure you’re building a scalable business

Timestamped Highlights

01:45 Sean’s Amazon journey and what time spent in warehouses taught him about customer pain points

04:14 The moment he saw the same issues plaguing both small and nine-figure sellers — and spotted an opportunity

07:37 How becoming a CEO forced him to rewire his focus beyond sales and build internal infrastructure

12:18 Why having a technical co-founder was non-negotiable — and how AI tooling is changing that equation

15:18 The tough decision to leave Amazon and how he measured risk versus regret

17:59 Learning to let go and trust others with the sales process while still staying close to customers

Memorable Moment

“Talk to the people that would actually buy your product. Measure the pain point. If it’s a one or two out of ten, it’s probably not worth building. If it’s a nine or ten, and they’ll pay for it, now you have something.”

Pro Tips

Validate early and price with intention. Don’t just ask if someone would use your product — ask exactly what they’d pay for it. Those conversations can save months of wasted build time.

Call to Action

If this episode resonated, share it with a friend who’s considering the leap into entrepreneurship. Follow the show for more conversations with founders, operators, and tech leaders building the next generation of companies.