08 / 08 · built · 2026 · Hackathon build
Agnes
Supply chain optimization engine for CPG companies: margin gains that stay EU-compliant.
- Role
- Full-stack: Spring Boot backend, wave-based optimizer orchestration, Claude integration, compliance engine, React frontend, graph visualization.
- Stack
- Java · Spring Boot · React · Vite · SQLite · Claude
- Links
- GitHub
Agnes acts as an AI supply-chain analyst for consumer packaged goods companies. It connects to a read-only SQLite database of ingredients, suppliers, products and companies, then uses Claude as its reasoning engine to find margin-improvement opportunities across four optimizer types: substitution, consolidation, reformulation and complexity reduction.
Optimizers run in waves on Java virtual threads. Wave one, substitution and complexity, runs in parallel; wave two, consolidation and reformulation, follows using wave one's results. Every finding then passes a two-stage compliance pipeline: a pre-filter auto-approves low-risk findings without an extra model call, and the remainder are verified against EU 1169/2011 with market evidence in a single batched call.
The frontend includes an interactive graph explorer with three views, company to supplier, company to product and product to supplier, so teams can navigate the supply chain visually alongside the chat interface.