thesis · 2026
DRIP - LLM agents for collaborative work
A real-time collaboration platform where an LLM agent turns team conversations into shared, negotiable tasks and documents.
The agent has 20 typed tools and runs on OpenAI or Gemini behind a provider-agnostic layer. It acts autonomously when someone addresses it, and only proposes when it observes private conversations, because a mistake there costs more.
Every change it makes is recorded, traceable to its source and confirmed by a person. Under the hood: Next.js, React, TypeScript, Postgres with row-level security and triggers, and CRDT-based co-editing.
- Role
- Solo developer and researcher
- Team
- Advisor: Prof. Maristella Matera
- Stack
- Next.js 15, React 19, TypeScript, Supabase, Postgres, Liveblocks, Yjs, OpenAI, Gemini, Vitest