Work With Me
Selectively available for short-term and contract projects.
I build production AI systems and the software around them. I take on a small number of projects alongside full-time work, so availability is limited.
AI Agents & LLM Systems
Agentic workflows with tool orchestration over MCP, retrieval-augmented generation, and the guardrails, retries and circuit breakers that keep them running in production.
Voice Agents
Speech-to-speech voice assistants built on LiveKit with Amazon Nova Sonic, wired into real backend tools rather than scripted flows.
AI Inside an Existing Product
Adding chat, search, automation or agents to software you already have, without rebuilding it.
AIOps & Data Pipelines
Pulling monitoring systems behind one interface, with anomaly detection, forecasting, change-point detection and root-cause ranking.
Full-Stack Product Build
End to end: PostgreSQL schema, FastAPI or NestJS service, Next.js frontend, deployment and observability.
How it works
Discovery call
30 minutes to understand the problem, the constraints and whether I'm the right person.
Scope & plan
A written plan: what gets built, in what order, and what "done" looks like.
Build & hand over
Working software, documented, with the context you need to run it.
Systems delivered
Includes contract work and systems delivered in employed roles.
Legal Action Media
United States, contractOwned the AI layer across a US personal-injury law firm's product suite: LangGraph/LangChain pipelines with PostgreSQL checkpointing for resumable claim-intake and document processing, plus a FastAPI service orchestrating OpenAI Agents SDK agents with LiteLLM multi-provider routing.
National telecom — AIOps console
anonymisedSix monitoring systems behind one connector SDK, with anomaly detection (isolation forest, random cut forest, seasonal), forecasting, change-point detection and root-cause ranking.
Enterprise RAG platform
anonymisedMulti-tenant retrieval on AWS S3 Vectors with Bedrock Titan v2 embeddings, one isolated vector index per tenant, over a 35-table schema.
Fashion-tech AI generation
anonymisedGoogle Gemini orchestration producing garment renders and structured specification tables, with per-call cost metering and a centralised retry layer.