Overview
What this challenge is about.
You receive 20 historical patent applications with the firm's own prior-art memos as ground truth. Design and build a 3-agent system: (a) Searcher — issues queries to a patent-search API (mocked) and a web search, (b) Reader — reads candidate documents and extracts claim-relevant passages, (c) Synthesizer — drafts the prior-art memo following the firm's house style. Compare against a single-agent baseline on memo quality (rated by 2 attorneys on a 1-5 scale across 4 dimensions). Write the 4-page architecture memo and a separate 1-page risk note for the managing partner.
The Brief
What you'll do, and what you'll demonstrate.
Determine whether a multi-agent system produces better first-pass prior-art memos than a single-agent baseline, on the firm's own historical cases.
Earning criteria — what you'll demonstrate
- Design multi-agent systems with explicit role decomposition
- Benchmark multi-agent vs. single-agent on a real task
- Apply human-rater evaluation to LLM-generated long-form outputs
- Communicate agent risk to a non-technical legal audience
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
AI Engineer
Designing and shipping a multi-agent system for a real legal customer is the kind of project that lets an AI engineer specialize into vertical AI roles.
This challenge sharpens
- multi-agent-collaboration
- llm-agents
- retrieval-augmented-generation