# PES Solutions

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Adewole Scott](https://cerebralvalley.ai/u/Pes_solutions)
- **GitHub:** https://github.com/adewolescott/agriroute-ai
- **Demo video:** https://youtu.be/CZeQw7L-XYY
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4399

In emerging agricultural economies, over 40% of fresh produce spoils before reaching regional wholesale aggregation points. Smallholder farming cooperatives and freight drivers face severe bottlenecks: illegible handwritten paper waybills, absence of standardized objective quality grading at rural farm gates, and completely fragmented logistics with zero cold-chain visibility. By the time produce reaches major urban markets, high spoilage rates lead to drastic financial loss for farmers and inflated food costs for consumers.

AgriRoute AI solves this post-harvest crisis by bridging Meta's state-of-the-art multimodal vision and audio intelligence directly with a deterministic geospatial routing infrastructure.

The platform automates a previously impossible manual workflow:

Multimodal Intake: A driver or local cooperative officer uploads a crate photograph, a handwritten waybill slip, or records an informal spoken trader voice note detailing cargo origin and transport urgency.

Zero-Shot Vision Grading & Segmentation: Segment Anything Model (SAM 3.1) isolates the produce crates from messy field backgrounds, and Meta Muse Spark 1.3 performs zero-shot visual defect evaluation—detecting mechanical bruising, skin discoloration, and ripeness indices.

Autonomous Manifest Generation: Using structured Pydantic schema decoding, Muse Spark 1.3 extracts crop varieties, normalizes imperial or informal crate measurements into metric tons, and computes ambient shelf-life limits. If shelf life is under 36 hours, it deterministically flags the cargo as CRITICAL with mandatory cold chain.

Spatial Dispatch & KNN Routing: The normalized manifest and farm coordinates are dispatched to a compiled Go microservice paired with PostgreSQL/PostGIS. The engine executes K-Nearest Neighbor (KNN) spatial matching against cold-chain storage facilities, calculating haulage distance, truck transit ETA, and capacity constraints.

Operator Dispatch Console: The Next.js 15 dashboard visualizes real-time economic impact metrics (cargo value preserved, spoilage risk mitigation percentage), automated inspection logs, and an interactive Leaflet corridor map connecting the farm gate to the matched cold-storage hub.

AgriRoute AI shifts AI from passive conversational chatbots into active, verifiable infrastructure that safeguards food security and farmer incomes.

## More from Meta Global AI Developer Hackathon

- [CASER](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4396)
- [JorgeosorioDev](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4397)
- [MusePilot](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4398)
- [CircularOS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4400)
- [Curato](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4401)
- [jaguar](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4402)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4399. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt