Kepler Innovation & Technology Culture

How does your team stay ahead of emerging technology trends while scaling fast?

A lot of it comes down to the people and networks around us. We’ve built strong relationships across the engineering and startup community, and those conversations are honestly one of our best sources of signal. Talking to other founders, engineers and operators who are deep in the same problems tells you more than any newsletter will. We also give people real autonomy to explore. If someone on the team wants to try a new tool or technology, they have the budget and the room to go do it.

There’s also a common assumption that staying on top of new technology gets harder as you scale, but we’ve actually found the opposite. Staying ahead is one of the things that enables us to move fast. AI has been a big part of that shift. Developers with strong fundamentals can ramp up on a new technology dramatically faster now. We can prototype and pressure-test something in days that used to take weeks, so the risk of evaluating a new approach is just fundamentally lower. You’re not committing to a massive rewrite to find out if something works. You can try it, learn fast, and make a real decision. That changes the whole calculus around adoption.

What recent product or feature are you most proud of — and what impact has it had?

Our finance product has been getting incredible feedback, and the thing I keep hearing from users is that they’re using AI for work they never would have trusted it with before. That’s the outcome, but the reason it works comes down to architecture. We built what we call an agent ontology: a system where AI handles reasoning and orchestration while deterministic code handles every retrieval and calculation. Every output is traceable cell by cell back to the exact page, table and line item in the source filing. Building that required solving some genuinely hard engineering problems. We run parallel sub-agents with domain-specific tool sets and validation checkpoints between stages, so data is verified before it ever reaches the output layer. Our APIs are strongly typed so that errors function as self-correction signals for the agents. And we built a discovery layer that maps semantic differences across companies and filing types so the system can query across entities without the orchestrator needing to know every edge case. It’s the kind of infrastructure that doesn’t exist off the shelf, which is what makes it a great problem for engineers who want to build something real.

How do you create a culture where innovation and experimentation are encouraged daily?

Innovation doesn’t come from setting aside special time for it. It comes from making experimentation cheap and fast enough that it’s just part of how people work every day. We invest heavily in the tools and infrastructure that make that possible. Better build tools and better debug environments, anything that shortens the loop between having an idea and knowing whether it works. A lot of that thinking comes from our time at Palantir, where we saw firsthand how much velocity you gain when you treat the developer experience as a first-class priority. We’ve also leaned into AI across our own workflows, from how we write and review code to how we handle meeting notes and project tracking. And we hire for it. We look for people who are excited about new technologies and naturally inclined to experiment. When you combine that mindset with low-friction tooling, you don’t need to mandate innovation. It just happens.