Etra Ventures, London · 2025 to 2026 · Lead developer
Foodie (opens in a new tab)
- Challenge
- Etra Ventures wanted a new consumer food-discovery product, built from nothing by a small team and taken live: video-heavy, location-dependent, and relying on two outside specialists, a data scientist's recommender and a Python video pipeline, whose work had to land inside one coherent product.
- What I built
- I led it from the first commit. The app is Next.js 16 and TypeScript on Supabase, with auth, row-level security and migrations in place before there were users to protect. The heavy work, video extraction, embeddings, quality checks and Places sync, runs in a separate Fastify worker on AWS ECS, provisioned with Terraform, so a slow pipeline job never slows the app. I integrated the recommender and the video pipeline through that worker and webhooks, and instrumented the whole flow in Amplitude so the team could see what people actually did.
- Outcome
- Live at foodie.app. A zero-to-production consumer app with its own backend, where the risky parts, ingestion, ML and third-party data, are isolated from the part users touch. That is the shape I give messy products.
- Next.js 16
- TypeScript
- Supabase
- Fastify
- AWS ECS
- Terraform
- Amplitude
Live: foodie.app
