V7

Most AI demos are impressive for 30 seconds. The hard part is making it actually work on messy, real-world data at scale.

I joined V7 as an AI Applied Research Engineer intern in London. V7 builds Go, an AI platform for automating document workflows with LLMs. Think parsing, extracting, and answering questions over messy real-world documents at scale.

Excel Q&A with Citations

Built an entire Excel Q&A workflow from scratch. Ask any question about any spreadsheet and it responds with a cited answer, linking you to the exact cell. No hallucinations, no vague answers. Every claim linked back to its source. Shipped to production.

AI Agents That Build AI Agents

Using LangGraph, I built multi-agent systems where AI agents could create, configure, and orchestrate other AI agents. Also shipped to production. The LangGraph for this was a nightmare.

LLM Benchmarking

Automated V7's LLM benchmarking. Spoke with LLMs non-stop and learned a lot (too much?) about PDFs.

Thursday pub nights with the team were one of the highlights of my summer. Cracked group of people.