Hey, I’m Robert Gumeny, an AI-Native Product Engineer in Bellevue, WA.
When I’m not at a keyboard I like to spend as much time outside as possible: hiking, camping, kayaking, and backpacking up and down the West Coast of the US with my wife and dog. I am learning how to draw and play the ukulele, and I feel that it is more important than ever to take time to unplug and have some analog hobbies. I read a lot, cook as often as I can, and care deeply about the climate and sustainable agriculture.
What I’m building
- 01
AKG
in-progressA portable, zero-setup binary graph format for agent memory. Compact, file-based, and built for embedding directly into agentic workflows and AI products.
- 02
Helm
early-accessHelm helps professionals own their career story by building a compounding career memory around their authentic experiences and achievements, one that grows richer with every interview, job application, and salary negotiation.
- 03
Agents Playing Poker
in-progressA research harness where AI agents play heads-up no-limit Texas Hold'em to evaluate memory strategies, proving that AKG-backed agents maintain a strategic edge while cutting inference costs by 90%.
- 04
doug
in-progressA lightweight Go CLI that acts as a deterministic orchestrator for coding agents, supporting the full agentic software development lifecycle with a built-in knowledge base that drives up to a 95% cache read rate.
- 05
robertgumeny.dev
liveThis site — a personal portfolio, blog, and home base built with Astro and deployed on Netlify.
About
Most AI products are built backwards, with the model at the center and everything else bolted on after. My philosophy is that the core of a great AI product is a deterministic context engine built by a team that understands the user’s workflow well enough to deliver the right context, to the right agent, at exactly the right moment to solve their hardest problems. Figuring out what the user actually needs is still the hardest part, and it matters more than ever when there’s an LLM in the stack.
My hot take: AI is a tool, not a strategy. The best AI-powered products are ones where the AI quietly handles the right narrow slice of work and everything else is solid, boring, deterministic software. When that’s the starting point, the infrastructure gets cleaner, the user experience gets better, and teams stop burning valuable runway chasing maxing out their tokens on the latest frontier model.
I strongly believe this is one of the most exciting times to be building software, and I think we have a real responsibility to keep the core building blocks as open as possible. The benefits of these tools should be available to everyone, not just teams with enterprise budgets, and that’s a big part of why I build in the open.
Let’s talk.
I’m looking for my next role at an early-stage startup where I can work on the hard problems: memory that compounds, agents that don’t fail users, and evaluation systems for subjective quality at scale. I’m also open to consulting on the right projects, and helping busy founders and individuals get a handle on their own personal AI workflows. Always happy to connect with other builders and product thinkers working on ambitious ideas. If that sounds like you, let's talk!