Farpoint · Research Engineer Intern · 2025–26
Fabric
Multi-agent orchestration, subagents and context management for an LLM-powered IDE.
- LLM agents
- Developer tools
- Evaluation
Overview
Fabric is an LLM-powered agentic IDE. I owned the agent layer: how work is split across agents, how they talk, which tools they may use, and how context fits in the window — plus the evaluations that decided what shipped.
Contributions and results
- Ran the multi-model orchestration study behind Fabric’s published benchmark: 97.8% on Aider Polyglot at $4.95 against 98.7% at $26.79 for the frontier baseline.
- Shipped a six-tool subagent system and a test-first, RED→GREEN multi-agent DAG orchestrator with a live control dashboard.
- Designed the prepare → permission → execute tool lifecycle, and an MCP server that lets agents drive the test-and-break loop.
- Ran a controlled SWE-bench experiment on graph-augmented code retrieval: no measurable gain, so it was not shipped.