I build and fine-tune custom models — LLMs, vision and time-series — and bring them into production, on the edge and in the cloud. Grounded in PhD-level research, with a focus on regulated, privacy-sensitive data.
Custom models, trained for your problem and your data and taken all the way into production — not just prompt-wrapping around an API.
Train and fine-tune models — LLMs, vision and time-series — with clean data, training and evaluation pipelines. For when off-the-shelf prompting isn't enough.
AI for healthcare, finance and legal — where privacy, GDPR and on-prem / edge constraints are the norm. Including GDPR-conform document processing with custom SLM / LLM models. I design for these constraints from the start.
De-risk a novel ML approach before you build a team around it. I prototype, validate rigorously, and tell you honestly what will and won't work.
A few anonymised examples across edge ML, applied research and optimization.
Real-time biomechanical motion analysis with on-device and GPU-cloud inference. Built the full data, training and inference pipeline end-to-end and owned the technical direction.
MVP shipped in 7 weeks — secured pre-seed fundingResearch on latent-reasoning that improves frozen LLMs and vision models — and earlier, novel deep-RL architectures for sequential decision-making.
Fundamental AI ResearchReal-time indoor navigation plus routing and pathfinding across large buildings. Built on PhD research in dynamic routing, scheduling and resource allocation — directly transferable to logistics and operations.
From raw floor plans to live, routable mapsMost people are either solid engineers or researchers or domain practitioners. I sit at the intersection: 14 years of software engineering, a PhD in deep reinforcement learning (2× IJCAI), and hands-on production ML across regulated industries.
That mix is what lets me take an idea from research feasibility all the way to a shipped, maintainable system.
I go all-in on a clearly scoped problem — defined goals, a defined outcome. You get full focus and a clear endpoint, rather than open-ended hours. Remote, with on-site where it helps.
I'd rather be the engineer you come back to again and again than one permanently parked on-site. A loyal, long-term relationship — built from many focused engagements.
Straight answers on what's feasible, the simplest thing that works, and no overselling. The data decides.
For recurring needs I can build, deploy and run a model as a service — you use it and pay for what you use, while I keep improving it.
If it's a fit, I'll say so. If it isn't, I'll tell you that too — no time wasted on either side.