The Tech Trek

Luke Fischer, cofounder and CEO of SkyFi, breaks down how earth intelligence is becoming searchable, and why that changes decision making across defense, energy, logistics, and agriculture.
You will hear how his path from Army special operations aviation to Head of Flight Ops at Uber shaped SkyFi’s product mindset, plus a practical look at what geospatial imagery and analytics can actually answer today.

Key Takeaways

• Networks are not nice to have, they are the fastest path to trust, hiring, and deals, especially in government and high stakes markets
• SkyFi’s core unlock is access, making it possible to task satellites, pull history, and ask questions of the data, not just look at images
• Going commercial first can create a faster iteration loop, then government adoption follows once the product is battle tested
• The real product future is answers, not imagery, using natural language queries that return decisions grade insight
• Privacy is not only about resolution, it is also about who can buy data, screening, and compliance, because access is the real leverage point

Timestamped Highlights

00:47 Earth intelligence in plain English, task satellites, pull decades of history, ask questions like vessel detection or soil moisture
06:32 Why veteran resumes miss the mark, and how to translate leadership without goofy title inflation
10:44 The origin story, a broken buying experience in satellite imagery turns into SkyFi’s wedge
16:42 Selling into government, people game first, acquisition reality, and why patience is a feature
19:46 Use cases you will not expect, livestock behavior, barge counting, palm heights, mineral detection, and more
28:10 Where this is headed, ask a question about the world, get an answer, then move toward proactive intelligence

A line worth repeating

“Startups are the same thing, you are finding the right people with the right traits to solve these undefined problems in being comfortable with risk.”

Practical moves you can steal

If you are hiring, screen for comfort with ambiguity, not just pedigree, undefined problems are the job in high growth work
If you are selling, build your network before you need it, warm paths beat cold volume every time
If you are building product, shorten the feedback loop, commercial iteration can harden the product before slower cycle buyers adopt

Call to Action

If this episode sparked ideas for how data, defense, or AI driven analytics will reshape markets, follow the show and turn on notifications so you do not miss the next one. Also share it with one operator who makes high stakes decisions and would appreciate a clearer view of what is happening on the ground.

Anish Agarwal went from MIT PhD researcher to founding Traversal, an AI company building intelligent site reliability engineering agents for the enterprise. In this episode, he breaks down what it actually takes to lead an AI first company when your entire career was built inside a lab.
This is not your typical founder story. Anish never planned to start a company. He was on track to be a professor at Columbia when generative AI hit and rewired his trajectory. Now he is two years into the CEO seat, recruiting top talent away from high paying jobs, and building a product at the intersection of causal machine learning and agentic systems.
We get into the mechanics of that transition. How do you go from publishing papers to pitching investors? What does storytelling look like when you are convincing engineers to leave comfortable roles and bet on your vision? And what happens when you start a company without even having an idea?
Anish also tackles a question the AI space is wrestling with right now. Is a PhD becoming table stakes for building an AI first company? His answer is more nuanced than you might expect. It is not the degree. It is the training. Reading the landscape, navigating uncertainty, and evaluating models with scientific rigor. Those skills separate builders from everyone else.
Key Takeaways
The best AI founders are not chasing credentials. They are leveraging research instincts to read where models and architectures are heading, and that foresight creates real competitive edges.
Starting a company without an idea is not reckless if you have the right co founders. Anish and his team showed up to a WeWork every day and treated idea exploration like a research problem until the right opportunity clicked.
Storytelling is the most underrated leadership skill in technical companies. Whether you are recruiting, raising capital, or explaining your product to nontechnical buyers, packaging complexity into a clear narrative is what moves people.
Every decision as a founder is a bet, including the decision to do nothing. Viewing inaction as a strategic choice changes how you prioritize and how fast you move.
As AI writes more code, someone has to make sure it works in production. That gap between code generation and reliability is where Traversal lives, and it is only getting wider.
Timestamped Highlights
(00:36) What Traversal does and why AI powered site reliability engineering is a massive unsolved problem in enterprise software
(02:00) The moment generative AI changed everything and why Anish walked away from a career he loved
(08:43) How Traversal found its problem without starting with an idea, and the co founder dynamic that made it work
(14:29) The real advantage of a PhD in AI and why it has nothing to do with the letters after your name
(19:49) Advice for PhDs entering the job market on how to position research experience so hiring managers actually get it
(20:29) Two years into the CEO role, what Anish wishes he had known and the skills that matter most for early stage founders
Words That Stuck
"If AI is writing your code, it has to fix it too. And right now it is only writing the code."
Founder Playbook
Pick a problem that sustains you for decades. Anish looks for problems that keep getting more complicated because that is where long term value compounds. If the problem has a ceiling, your company does too.
Treat recruiting like a core product skill. Painting a compelling picture of the mission is not a nice to have. It is the engine that pulls exceptional talent away from safe, well paying jobs.
Think of everything as a series of bets. Fundraising, hiring, product decisions, even waiting. Inaction is a bet too. Once you see it that way, you stop overthinking and start moving with intention.
Subscribe to The Tech Trek wherever you listen. If this one hit home, share it with a founder or tech leader navigating their own leap. Follow the show on LinkedIn for more.

Harry Gestetner built a creator economy platform in college, sold it, and walked away. Then he did the one thing nobody expected. He jumped back in and started building hardware.
In this episode, the founder and CEO of Orion (a sleep tech company making smart mattress covers) sits down to talk about what really happens after an exit, why most founders can't stay away from building, and what changes when you go from software to physical products.
Harry shares what surprised him about the acquisition process, how he thinks about evaluating new startup ideas, and why he believes hardware is "life on hard mode." He also gets into the mental side of founding, from managing stress to staying sharp when everything feels uncertain.
What You'll Walk Away With
Going through an exit sounds like the finish line, but Harry explains why it's actually a reset. You trade ownership and freedom for financial security, and at some point, most founders start craving the creative control they gave up.
Not every idea deserves your time. Harry talks about running new concepts through a "disqualification period" where you actively try to poke holes before committing. The ones that survive that process are worth going all in on.
Hardware changes the game. Software lets you pivot fast. Hardware gives you 18 month product cycles, inventory headaches, and supply chain complexity. Conviction has to be higher before you start.
The best startup ideas come from problems you and your friends actually have. If enough people share that problem, you've got a market.
Knowledge compounds across startups. Harry compares the founder journey to an elastic band. Once you've been stretched, you never go back to your original form. Every challenge you survive makes the next one more manageable.
Timestamped Highlights
[00:34] What Orion actually does and how it makes six hours of sleep feel like ten
[03:01] The emotional arc of an exit that nobody talks about, from relief to restlessness
[05:34] How Harry evaluates startup ideas and why he uses a disqualification process
[09:30] Why building hardware is "life on hard mode" and what made him take it on anyway
[10:39] The elastic band theory of founder growth and why learning compounds over time
[15:49] His advice for early career founders: pick one thing and go all in
Words That Stuck
"As a founder, you're sort of like an elastic band. The more you get stretched, you never go back to the original form."
Tactical Takeaways
Run every new idea through a disqualification period. Actively look for reasons it won't work before you commit. The ideas that survive that scrutiny are the ones worth building.
Build around problems you personally experience. If your friends share the same frustration, there's a good chance others do too. That's your market signal.
If you're going to start something, go all in. Stop hedging across multiple projects. Pick one idea and dedicate yourself to it completely until it works.
Keep Up With The Show
If this episode hit home, share it with a founder or someone thinking about taking the leap. Subscribe wherever you listen so you never miss an episode. And connect with us on LinkedIn for more conversations like this one.

Behnam Bastani, CEO and cofounder of OpenInfer, breaks down why the last two years of AI feel explosive, and why the next wave is not chat, it is action at the edge.
We get into always on inference, what actually forces compute to move closer to the data, and the missing layer that makes edge AI scale: the Android like infrastructure that lets devices collaborate instead of living in silos.

Key takeaways

• The hype spike is real, but the runway is decades, it took compute, sensors, and communication protocols maturing over generations to unlock this moment
• AI is shifting from conversational to actionable, which means continuous, always on inference becomes the norm
• Edge wins when cost, reliability, and data sovereignty matter, cloud and edge will coexist, but the workload placement changes
• The biggest bottleneck is not just silicon, it is the infrastructure layer that makes building and deploying across devices easy, plus a shared fabric so devices can cooperate
• Adoption is as much a human story as a technical one, this shift lands faster and broader than previous tech transitions, so anxiety is predictable and needs real attention

Timestamped highlights

00:38 OpenInfer’s mission, intelligence on every physical surface, and why collaboration matters
02:07 Electricity as the earlier revolution, intelligence as the next kind of power, and the control problem
05:54 Where we really are on the maturity curve, early products are here, mass adoption and safety take time
08:31 When the device boundary disappears, it stops being you versus the agent, it becomes one system
11:04 Always on inference, and the three forces pushing compute to the edge: cost, reliability, data sovereignty
14:40 The Android moment for edge AI, why the operating system layer unlocks developers, apps, and adoption

A line worth replaying

Those are going to be the three pillars that really enforces that edge and cloud are going to live together.

Pro tips for builders

• If your product needs real time decisions, design for intermittent networks from day one, reliability is not optional
• Treat data sovereignty as a product feature, not a compliance afterthought, it is becoming the moat
• Push for interoperability early, the fabric that lets devices share the right data is what makes edge feel seamless

Call to action

If this episode helped you rethink where AI should run and what it takes to ship it in the real world, follow the show and share it with one builder who is working on edge, robotics, devices, or applied AI.

Building data capability from zero is not a tooling problem, it is a trust and prioritization problem. In this episode, Laura Guerin, Head of Data and Data Science at Bevi, breaks down how she goes from blank slate to real business impact, without getting trapped in endless plumbing or endless meetings.

Laura shares how she runs an early listening tour, prototypes value before asking for bigger investment, and decides when to hire scrappy generalists versus specialists. We also get practical on AI, where it helps, where it is unnecessary, and why quality data and a clean semantic layer still decide whether anything works.

Key takeaways
• Start with business priorities, then map data work to the actions and outcomes leaders actually care about
• Prototype the end deliverable fast, even if the backend is duct tape at first, then scale after stakeholders see value
• Use cases first for AI, most problems do not need AI, but the right problems can see real acceleration
• Early teams win with adaptable generalists who can wear multiple hats across data, analytics, and data science
• Trust is a shared responsibility, build reliability, then create a culture where users flag weirdness quickly

Timestamped highlights
00:44 Bevy explained, smart bottle less dispensers and why the business context matters for data priorities
02:01 The listening tour playbook, exec alignment, stakeholder map, and using AI to synthesize themes into a SWOT
04:00 The MVP reality, manual prototypes to prove value, then the conversation about scalable pipelines
06:33 AI without the hype, use cases, when AI is not needed, and two examples with clear business impact
09:22 Hiring from zero, why generalists first, the data analytics data science spectrum, and the personality traits that matter
14:21 Self service reimagined, Slack as the interface, semantic layer and permissions, and how to keep a single source of truth
20:19 Keeping trust when things break, checks and balances plus a shared responsibility model
22:39 Making innovation real, baking it into expectations so the team has time to learn and test new approaches

A line worth stealing
Data on its own is not typically a priority. It is more about the action or the impact that comes out of the data.

Pro tips
• Run a structured listening tour early, capture themes, then pick two or three priorities you can deliver quickly
• Show the business an MVP output first, then use that proof to justify the unglamorous backend work
• Treat AI like any other tool, define the problem, validate the use case, then confirm the data quality inputs

Call to action
If you are building analytics, data products, or AI inside a growing company, follow the show and subscribe so you do not miss the next operator level conversation. Share this episode with one leader who is asking for data outcomes but has not funded the foundation yet.