ML research, AI engineering, and production software in one person, built on 12 years of writing code and doing computer science.
At LinkedIn, I built the forecasting model behind a nine-figure hardware allocation.
The rigor of a researcher, the output of a senior engineer, and the judgment of a CTO. Four ways to put them to work:
Models that hold up in production, not just in a notebook. I design and build ML systems: LLM applications, RAG pipelines, agent harnesses, and custom models, plus the evaluation, monitoring, and infrastructure that keep them reliable once real users and real money depend on them.
For problems with no off-the-shelf answer. I frame the question, survey the literature so your team doesn't have to, design the experiments, and develop new methods where existing ones fall short. Then I turn the result into code your team can run. This is the work behind my Cornell PhD, a NeurIPS workshop paper, and an AMD patent filing.
12 years of writing software, from a C++ system for training and deploying ML models in HPC environments at AMD to full-stack production platforms. I built ArcScore Inc.'s platform end to end: front end, back end, data pipeline, CI/CD, and staging and production environments with observability.
Senior technical judgment without a full-time hire. As a fractional CTO I set technical direction, make build-vs-buy calls, and lead engineering teams. I also embed with a team for a stretch, in your repo and your code review, and for the right founders I take on a technical cofounder role.
One person from the research question to the production system. Nothing gets lost between the people who understand the math and the people who ship it, because they're the same person.
I learn the business problem first. If the answer isn't known yet, I scope the research question and decide up front how we'll measure success. Then I design the system: how the pieces fit together, how they scale, and how they fail.
No six-week discovery phase. A working prototype comes early, measured against the success criteria we set, so we're debating evidence instead of a slide deck.
I harden what works into production: tests, observability, documentation, and infrastructure you can measure and scale.
Documentation, runbooks, and training until your team can maintain and extend it independently. I'll stick around if you want me to, but you won't need me to.
Ten minutes is enough to tell whether I can help. If I'm not the right fit, I'll tell you.
Book a Free 10-Minute Discovery Call
I'm a machine learning researcher and a systems builder. I've spent 12 years writing software and doing computer science, and I hold a PhD in Computer Science from Cornell (2026). I'm a named inventor on AMD's patent filing on accelerating physics-based simulations with AI. The research matters because I ship it.
My dissertation developed practical algorithms for multi-step lookahead Bayesian optimization: deciding what to try next when every experiment is expensive. Along the way I did research at Adobe, IBM, AMD, and LinkedIn, on natural language interfaces, analog AI hardware, uncertainty quantification for scientific ML, and forecasting. I read the papers, and I write the production code.
I can whiteboard your system architecture, write the code myself, deploy it to production, and explain to your investors why it matters. Since October 2025 I've been doing exactly that as founding engineer at ArcScore Inc.: I architected and built the platform end to end, designed the core scoring and valuation methodology with two other modeling leads, and I own significant portions of the application, data, infrastructure, and roadmap.
The thread running through my work is ML applied to the economics of expensive, constrained resources: forecasting a nine-figure hardware allocation at LinkedIn, accelerating physics-based simulations at AMD, pricing athlete value at ArcScore. If your hard problem lives where models meet infrastructure, that's my lane.
I take on consulting, fractional technical leadership, and selective founding work through this site. I'm also open to the right full-time research or senior engineering role; my resumes and full CV are on my work page.