The long version.
Home is the resume. About is the worldview — how I work, and the principles I'd defend in a design review.
I'm a tech lead and platform engineer based in New York. My day job is owning the infrastructure that ML teams ship on — Kubernetes, inference servers, CI/CD, and the developer-experience tooling that collapses a 40-minute “set up your laptop” into a single command. I led a 3-person platform team at Technergetics through 2024 — sprint planning, hiring input, and the architecture calls — and now carry the domain as tech lead.
Before that, I trained as an applied mathematician and data scientist. That side of the brain still drives most of what I find interesting: tiny models that beat big ones, retrieval policies that learn, the unreasonable effectiveness of well-placed caching.
I also run a self-hosted Kubernetes homelab with full GitOps, SSO, observability, and CI/CD, and I'm building Vivazo — a Rust game engine targeting WASM. Both are how I keep the systems parts of my brain sharp on weekends.
Keep production boring
Production should be unsurprising. Save the cleverness for the parts that move the needle — the inference path, the retrieval policy, the model contract. Everything else is plumbing, and plumbing should be conventional.
Make it one command
Anything an engineer types more than three times deserves a script. Anything a team types more than three times deserves a platform.
Mirror prod in dev — actually
Half-faithful local stacks are worse than no local stack. If staging passes and prod fails, your dev environment lied. The hard part is bringing up the real dependencies together, because bugs live at the seams between services.
Measure the path, not the model
Most ML wins come from the retrieval policy, the eval harness, the cache hit rate — not the LLM. A small model in the right place beats a big one in the wrong place.
M.S. Data Science
Focus on statistical learning, NLP, and large-scale ML systems.
B.S. Computer Science & Applied Mathematics
Dual major. Research support specialist (2019—2020) before grad school.