You have an idea or a prototype.
You want to bring it to market and support real users. I design an architecture that can evolve with the product and connect the parts that make it work.
The recruiting platform case studyWorking together.
Software and AI, from the first technical decision to production. You work directly with me.
A product to build or take further.
I work across the whole system: backend, interfaces, real-time data and infrastructure. I use AI where it solves a concrete problem.
A manual process is slowing you down.
I identify what to automate and where human judgment matters. I build integrations and pipelines with LLMs and agents, with checks on both the input data and the results.
The medical reports case studyFrom the first call to handover.
We start with a 30-minute call, with no obligation. You tell me about the problem, goals and constraints; we look at what already exists and work out whether I can help.
A written proposal.
We set out the scope, timeline, price, deliverables, responsibilities and acceptance criteria. We can work on a fixed-price project or an ongoing engagement with an agreed monthly allocation of hours.
Reviewing progress every week.
I compare technical options and use focused prototypes to test unknowns that could change the project. I work in short iterations: each week, we review running software and open decisions. If the scope changes, we discuss it before changing direction.
Release and handover to your team.
I handle release, including observability and error handling. I deliver code, tests, CI and documentation, including technical decisions. Repositories and infrastructure stay under your control. I help your team take over, and we agree on how to support the system as it evolves.
The tools I usePython, TypeScript, AI and infrastructure
I’ve been shipping AI systems since GPT-3, before ChatGPT. I choose tools around the problem and the people who will maintain the system.
- Primarily Python
- Django and real-time data pipelines, with the rest of the stack chosen to fit the product rather than forced into a single framework.
- TypeScript, React, Next.js
- Clean Tailwind CSS interfaces, from UI implementation through deployment.
- LLMs and multi-agent systems
- Document extraction, embedding and reranking, agents and open-source models hosted on NVIDIA GPUs. Where it makes sense, self-hosting gives you control over data and infrastructure and helps estimate running costs.
- Docker and Terraform
- Docker and Docker Compose on every project, with Terraform for versioned, reproducible infrastructure.