The Tech Trek

In this deep-dive episode, we explore what it truly means to be "AI-native" versus bolting AI onto existing products. Abhay Mitra, CTO of Nirvana Insurance, shares how his team is building industry-specific AI models to transform the $800B+ commercial insurance market, starting with trucking—one of the most complex and painful sectors in insurance.

From telematics data platforms to fine-tuned underwriting models, discover why commercial insurance might be the perfect proving ground for AI and how a data-first approach is creating unfair advantages for startups competing against century-old incumbents.

Key Takeaways

🎯 AI-Native vs. AI-Enhanced: Know the Difference

AI-Enhanced: Adding chatbots and customer service automation to existing workflows

AI-Native: Building core business logic, pricing, and underwriting around AI models from day one

The key differentiator: domain-specific data and expert annotations that create defensible moats

📊 Data is the New Competitive Moat

Quality beats quantity: Having "heaps of data" means nothing if it's not structured and usable

The real challenge: Correlating data across 20-100 different legacy systems

Version control for AI: You need to remember what models and rules applied at what time to properly train new models

🚛 Why Commercial Insurance is Perfect for AI

10-15x more complex than personal insurance with premiums to match

Highly varied customer profiles that resist traditional automation

Perfect storm: Complex data + high-stakes decisions + massive inefficiencies = AI opportunity

🏗️ Building AI-Native Engineering Teams

Hire for data expertise first, AI expertise second

Invest 5x more time in data quality and expert annotations than traditional SaaS

Focus on reliability and production-readiness, not just impressive demos

💰 The Startup Advantage Over Legacy Players

Legacy companies have data but can't correlate it effectively across systems

Modern data infrastructure beats decades of accumulated technical debt

Speed of iteration trumps size of existing datasets

🕒 Timestamped Highlights:

00:00 – 02:18: Intro to Nirvana Insurance and choosing to tackle the hardest problems in commercial insurance.

03:22 – 06:40: Why off-the-shelf AI isn’t enough and how domain-specific modeling gives Nirvana an edge.

07:28 – 09:55: Defining what's core IP vs. commodity tech when building AI solutions.

10:28 – 13:45: Why commercial insurance is a perfect fit for AI—high complexity, high stakes.

17:10 – 20:13: The difference between data-first and AI-first engineering orgs.

20:58 – 23:59: Why legacy insurers struggle to operationalize their data despite decades of collection.

25:09 – 27:26: What customers actually care about—better outcomes, not flashy tech.

💬 Quote:

“Before AI, this wasn’t even possible. You just couldn’t bring that level of nuance to each individual business. But with these new capabilities, insurance can finally become a tool for safety—not just cost.” — Abhay Mitra

What's Next?

Enjoyed this deep dive into AI-native insurance? Share this episode with your network and subscribe for more conversations with CTOs and engineering leaders building the future of regulated industries.

Questions or feedback? Drop us a line—we read every message and love hearing how these insights are helping you build better products.

In this episode, we dive deep into the evolving relationship between engineering and product with Pranab Krishnan, CTO of Zeal - a payroll and payments platform for staffing companies. We explore how the traditional boundaries between engineering, product management, and customer interaction are dissolving, especially in the age of AI. Pranab shares insights on building a product-centric engineering culture, the concept of "shifting left," and how AI tools are reshaping the skills engineers need to succeed.

Key Takeaways

🎯 Everyone Should Be a Product Person

The most successful startups foster a culture where engineers, designers, and even operations staff think like product managers and maintain direct connections to customer needs.

🤖 AI as the New Abstraction Layer

Just like TypeScript abstracted JavaScript complexity, AI will become another abstraction layer. The future belongs to those who master orchestration, architecture, and agency - not just coding.

🚀 The Flat Organization Future

Teams will become leaner and flatter, with higher expectations for product surface area. Instead of hiring more engineers, companies will be expected to build more comprehensive platforms with the same team size.

⚡ Shift Left Philosophy

Engineers moving closer to business problems and customer interactions, while designers and other roles also expand their responsibilities into traditionally separate domains.

🏗️ Core vs. Edge Development

In regulated industries like fintech, maintain bulletproof core systems while moving fast on user-facing features and interfaces.

Timestamped Highlights

[01:26] The CTO Evolution Journey

[03:44] Building vs. Learning Product Skills

Pranab discusses whether product management is learnable or innate, emphasizing that everyone approaches it differently - some from operations, others from technical backgrounds.

[06:12] The AI Evolution Question

Discussion on whether AI represents an evolution of software engineering or a fundamental paradigm shift away from core coding skills.

[07:21] AI as Abstraction

"My thesis on this is that everything is an abstraction... We are going to see AI becoming abstraction. The skills that I think people will need over the next five to 10 years is... orchestration... and agency."

[10:49] The Backlog Problem

Exploring what happens to product backlogs when engineers can produce more through AI assistance, and the potential for engineers to become natural problem-solvers with more time.

[15:15] Magic Patterns Tool Discussion

Real-world example of AI design tools that allow rapid UI iteration and prototyping with simple prompts.

[21:29] Expertise and AI Questions

"You can judge expertise by the types of questions people ask. And I think these tools... it requires a technical person to ask those questions, because you're not gonna know the nuance of if the answer's correct or not."

[23:38] The Future Hiring Landscape

Prediction that while teams will initially hire fewer engineers, expectations for product complexity will increase, eventually balancing back to similar hiring needs.

[25:13] The Data Advantage

"The big AI company that's going to do this well is most likely going to be the one who has the most data about you. So OpenAI is already poised... I would not be surprised if OpenAI builds its own Netflix, its own web flow, its own e-commerce."

Featured Quote

"Intelligence is on tap, but agency is the core of capitalism. Agency is going to be even more important as intelligence is more easily available to us."

— Pranab Krishnan, referencing Gary Tan

Tools & Resources Mentioned

• Magic Patterns - AI-powered design tool for rapid UI creation
• Zeal - Payroll and payments platform for staffing companies
• Claude Sonnet 3.7 - Referenced for physics simulation capabilities
• Y Combinator philosophy on hiring

Like this episode?

Share it with fellow tech leaders and subscribe for more insights on the intersection of technology, product, and business strategy.

In this episode of The Tech Trek, Vinayak Kumar shares how his team at Lynx strikes a practical balance between innovation and efficiency in the heavily regulated healthcare and finance space. He explains why innovation shouldn’t be forced, how to avoid the "tech in search of a problem" trap, and why pattern-driven execution helps startups scale faster without compromising flexibility.

🔑 Key Takeaways:

Innovation Should Be Embedded, Not Mandated

Innovation at Lynx happens organically—it's not about buzzwords, it's about solving real problems with the right tools.

Avoid “Technology in Search of a Problem”

True innovation stems from understanding the business problem first, then choosing a tool—not the other way around.

The Power of Reusable Patterns

Solving a problem once and codifying the solution into repeatable patterns has helped Lynx grow quickly and stay lean.

Fungibility in Teams Is Critical

Developers are encouraged to work across tech stacks to increase agility and reduce dependency on specialized roles.

🕒 Timestamped Highlights:

[02:55] – Why innovation must be cultural, not a KPI

[05:38] – Real-world example of choosing technology based on a business problem

[07:59] – The trap of adopting AI without a clear use case

[09:49] – Defining and leveraging “cookie cutter” solutions without sacrificing flexibility

[13:10] – A rigorous, fast-paced tech evaluation process in regulated industries

[16:41] – How Lynx builds team flexibility through cross-functional experience

[19:44] – Using agentic AI to automate non-obvious internal tasks like production issue research

💬 Featured Quote:

“We don’t talk about innovation—we just solve problems. And when you do that every day, innovation takes care of itself.”

In this episode, Amir speaks with Ameya Brid, Global Director of Data & Analytics at Invista, about the maturation of GenAI conversations in the enterprise. They dive into the shift from hype to implementation, real-world challenges like data quality and change management, and how composable architecture is helping organizations adapt to rapid innovation cycles.

🔑 Key Takeaways

From Hype to Value: GenAI conversations are moving beyond experimentation into outcome-driven initiatives—but most companies still struggle to define measurable KPIs.

Top Barriers to Scale: Poor data quality, fragmented systems, unclear use cases, and skills gaps continue to stall enterprise GenAI efforts.

Composable > Monolith: Modular, API-driven architectures provide agility to swap components as the tech rapidly evolves.

Change Management Rebooted: Adoption now means embedding insights directly into workflows—not just “viewing reports.”

Upskilling is Social: Peer-driven learning and internal documentation are outperforming formal training in the GenAI era.

🕒 Timestamped Highlights

00:00 – Introduction to Ameya and Invista’s work in manufacturing and chemicals

01:58 – How GenAI conversations have evolved over the past 18 months

03:52 – Marrying business outcomes with AI capabilities

06:04 – The five biggest barriers to GenAI implementation: use case clarity, data quality, skills gap, governance, and change management

11:53 – Managing constant tech evolution with composable architectures

15:02 – Data quality’s outsized impact on GenAI success

17:46 – Why CFOs must now invest in data quality

20:41 – Change management: From “read the dashboard” to “integrate AI into your workflow”

24:03 – Upskilling through shared learning and internal knowledge loops

💬 Quote of the Episode

"The cost of bad data today is far higher than it was 10 or 20 years ago—not just in decision-making, but in the process itself." – Ameya Brid

In this episode of The Tech Trek, Amir sits down with Andy Beam, CTO of Lila Sciences, to explore how AI is transforming the messy, serendipitous nature of scientific discovery into an engineered, scalable process. From automating lab work to accelerating the speed of breakthroughs, Andy explains why the future of science may be less about eureka moments and more about AI-driven iteration.

🔑 Key Takeaways:

Science as Engineering: AI enables science to move from a lucky break model to a systematic engineering process.

Scaling the Scientific Method: Pairing AI with experimentation platforms creates a feedback loop where hypotheses can be tested at unprecedented speed and scale.

Productivity Shift: AI copilots are redefining how scientists (and technologists) interact with their work, elevating humans to higher levels of abstraction.

Compounding Innovation: Once AI systems start discovering consistently, the rate of breakthroughs could go from decades to weeks—shifting timelines across industries.

⏱️ Timestamped Highlights:

00:00 – Intro to Andy Beam and Lila Sciences

01:00 – Why the scientific literature is a record of debate, not facts

03:09 – Science’s reliance on serendipity—and why that’s changing

04:55 – The power of scale in AI and what it means for discovery

06:15 – Andy’s personal shift in programming with AI copilots

08:41 – Will AI cause serendipity instead of waiting for it?

09:38 – The fungibility of speed and intelligence in research

11:47 – The challenge of change management in scientific communities

13:30 – What consumer adoption could look like in a future of constant innovation

💬 Quote:

“What we’re doing is taking the scientific method and scaling it with AI—so instead of waiting for Einstein, we build a million of them and run them 24/7.” – Andy Beam

In this episode of The Tech Trek, Amir speaks with Alexander Schlager, founder and CEO of AIceberg, about how his company has tackled the AI talent shortage by partnering directly with universities. From building relationships with faculty to onboarding students into real-world R&D roles, Alex shares a unique, cost-effective strategy for hiring early-career tech talent and turning them into long-term contributors. It’s a compelling listen for anyone in emerging tech, hiring, or leadership.

🔑 Key Takeaways

Faculty Buy-in Is Crucial: AIceberg’s success hinged on close collaboration with university faculty, ensuring student recruits were well-prepared and supported.

Rethink Talent Pipelines: Instead of competing for senior AI engineers, they invested in training early-career talent—gaining loyalty and retention in return.

Process Over Pedigree: Success in junior hires wasn’t about academic brilliance alone—it required a willingness to follow processes and grow into professional environments.

Retention Through Learning & Ownership: Clear career paths, challenging problems, and the ability to own projects helped retain young talent even with lower initial salaries.

⏱️ Timestamped Highlights

00:30 – What AIceberg does: AI trust platform for monitoring AI interactions

01:39 – The challenges of hiring AI talent in a startup environment

03:17 – Why partnering with faculty made their hiring model work

05:42 – Managing overhead and coaching needs with junior hires

08:02 – Standardizing research and product pipelines with JIRA

10:24 – Who to contact when building university partnerships

11:50 – Why maturity and teamwork matter more than grades alone

14:43 – How AIceberg advises candidates to evaluate offers before accepting

16:49 – Documentation and redundancy reduce risks when junior hires leave

18:30 – From outreach to onboarding: a 3-4 day ramp-up process

20:18 – Fresh perspectives from new grads as a strategic advantage

💬 Quote

“Don’t underestimate the benefit of a fresh brain—students often approach problems in ways seasoned professionals might never consider.”

Wyatt Smith, CEO of UpSmith, joins Amir to unpack how agentic AI is transforming the skilled trades industry. From dispatch optimization to human-in-the-loop workflows, Wyatt shares a practical and visionary lens on how AI can solve deep productivity challenges, empower call centers, and proactively generate business opportunities. If you think AI only disrupts digital industries, this episode will make you think again.

🔑 Key Takeaways:

Agentic AI is unlocking productivity by automating repetitive coordination tasks—like technician dispatching—allowing humans to focus on higher-value interactions.

Skilled trades businesses already have rich data but need tools to surface and act on it proactively rather than reactively.

Selling AI into traditional industries requires proof points, tight business cases, and sensitivity to the human element.

AI augments, not replaces—freeing up people to do work they're best suited for, like nuanced customer engagement.

💬 Highlight Quote:

“Advances in technology automate tasks, not people… Machines do what they're best at so humans can do what they're best at.” – Wyatt Smith

⏱️ Timestamped Highlights:

00:38 – Intro to Wyatt Smith and UpSmith's mission in the skilled trades.

02:51 – Why dispatching the wrong tech to the wrong job is a billion-dollar coordination problem.

05:09 – The customer journey in home services—and where productivity breaks down.

08:54 – AI adoption challenges in the trades and how business owners evaluate new tech.

11:15 – Human-AI dynamics: skepticism, latency, and building trust with agentic systems.

13:49 – “AI creates more work”: how automation changes tasks, not headcount.

17:19 – How UpSmith trains agents like new hires with workflows and documentation.

20:31 – Personalization at scale: how agents remember details from 5 years ago.

23:20 – The future of call centers and human-in-the-loop automation.

25:49 – Wyatt’s contact info and closing reflections.

What separates a successful founder from the rest? In this episode, Harish Abbott—CEO and co-founder of Augment—breaks down how he repeatedly spots opportunity early, builds products customers actually want, and navigates the fast-moving world of AI without falling into the trap of chasing every shiny benchmark.

We explore how Harish’s team shadowed 60 logistics operators before writing a single line of code, why storytelling is a founder's most underutilized superpower, and how to know when it’s time to pivot—even if everything looks good on the surface.

Whether you're scaling your first product or figuring out what not to build, this conversation is packed with real-world insights you can apply today.

🔑 Key Takeaways:

Start with Pain, Not Product: Successful startups begin by deeply understanding real customer pain points, not by jumping into code or chasing tech trends.

Shadowing Over Selling: Harish’s team shadowed 60 logistics operators in the early days of Augment—prioritizing observation over assumptions.

Strong Opinions, Loosely Held: Founders must balance confidence in their vision with humility to pivot when data points to a better path.

AI ≠ The Product: In a world obsessed with benchmarks, remember: AI is a tool. The actual value lies in making things better, cheaper, or faster for users.

⏱ Timestamped Highlights:

00:32 – What Augment does: AI teammates for the logistics industry

02:48 – “Follow one path consistently” – Harish’s approach to serial entrepreneurship

05:57 – The importance of shadowing operators before writing code

11:21 – When is it time to pivot? Why usage data is often more telling than top-line growth

19:23 – Storytelling as a founder’s core job: how to get employees, investors, and customers on board

25:02 – The challenge of AI startup building today: chasing stability over shiny new benchmarks

30:10 – Avoiding the trap of benchmark chasing in AI product development

💬 Quote:

“The best founders are always seeking truth. That truth sometimes tells you to let go of the idea you love.”

In this episode of The Tech Trek, Amir speaks with Patrick Leung, CTO of Faro Health, about what it takes to lead an engineering organization through a transformation to become an AI-first company. From redefining the product roadmap to managing cultural and technical shifts, Patrick shares practical insights on team structure, skill development, and delivering AI-enabled features in a regulated domain like clinical trials. This is a must-listen for tech leaders navigating similar transitions.

🧠 Key Takeaways:

AI-First ≠ Just Using AI

Being AI-first means deeply embedding AI into the core product architecture—not just bolting on an LLM. It requires strategy, structure, and long-term thinking.

Build the Right Team Early

The biggest shift for engineering orgs is in people—getting the right AI talent onboard early, rather than doing it all yourself, is critical for momentum.

Upskilling Is Real—but Selective

Not every engineer will pivot to AI, but there’s room for involvement across UX, product, and front-end roles. Cultural fit and willingness to contribute matter more than title.

Data Engineering is the Unsung Hero

Most AI work today isn’t in model building, but in crafting clean, structured datasets. Investment here pays off exponentially.

⏱️ Timestamped Highlights:

00:00 – What Does It Mean to Be AI-First?

Patrick defines the term and outlines Faro Health’s mission to reduce the cost and timeline of clinical trials.

04:13 – Defining the AI Strategy

How they started with clinical writing as the first application of LLMs and why it was harder than expected.

07:54 – The Role of Change Management

AI introduces massive shifts; managing sponsor expectations and workflows is as important as the tech.

10:28 – Engineering Impact

How the roadmap changed and what it meant for full-stack vs. data science roles.

14:24 – Hiring vs. Upskilling

Why Patrick hired an expert to lead AI efforts and the balance between internal upskilling and external hiring.

16:43 – Competing for AI Talent

How startups can win top AI talent despite the lure of FAANG compensation.

18:58 – Team Culture and Opportunity

Creating space for engineers who want to jump into AI while maintaining alignment on startup needs.

21:07 – Realistic Upskilling Paths

From Coursera to immersive bootcamps—what actually works for engineers wanting to break into AI.

23:11 – If He Could Do It Again

The two things Patrick would do sooner: hire a dedicated AI team and build structured data pipelines earlier.

🔖 Featured Quote:

“If you're serious about becoming an AI company, you need to find someone amazing who's launched real AI products—and build a team around them.”

In this episode of The Tech Trek, Amir sits down with Sunita Verma, CTO at Character AI and former engineering leader at Google. Sunita shares how she’s transitioned from leading large-scale AI initiatives at Google to building novel experiences in a fast-paced startup environment. She dives into the mindset shift required to prioritize velocity over scale, how to lead AI-native product innovation, and what it means to be a female technical leader in today’s tech ecosystem.

🔑 Key Takeaways:

Shift in Leadership Mindset: At startups, leaders must prioritize velocity and innovation over scale, focusing on getting frictionless, AI-native products to market quickly.

AI Product Loop: Success comes from tightly coupling AI research with product development—shortening the feedback loop to create truly novel user experiences.

Female Technical Leadership: Sunita emphasizes the need for more women in senior engineering roles and shares how calculated risk-taking and mentorship shaped her journey.

Startup Clarity vs. Corporate Comfort: While startups offer focus and purpose, they also require deep ownership and rapid decision-making without the cushion of big-company resources.

💬 Quote:

“Focus brings clarity of purpose... but with that comes the pressure of knowing every decision deeply impacts the company.” — Sunita Verma

⏱️ Timestamped Highlights:

00:00 – Intro: Meet Sunita Verma, CTO at Character AI and former Google engineering leader.

01:52 – Google to Startup: Comparing work at Google with her current role at Character AI.

03:39 – Leadership Shift: Sunita’s take on building AI-native products from scratch.

06:21 – From Scale to Speed: Pivoting from optimization at scale to innovating with velocity.

08:12 – Product & Tech Integration: Creating tight feedback loops between AI research and products

10:01 – Closer to Engineering: Why Sunita enjoys being hands-on and deeply involved in compute management.

12:12 – Focus as a Double-Edged Sword: The simplicity and pressure of startup leadership.

14:00 – Female Engineering Leadership: The need for more women in senior tech roles.

16:02 – Career Advice: Why calculated risk and building a support network are key to long-term success.

19:14 – Leaving Google: Her thought process in taking the leap from a big brand to an emerging category leader.