-
01 — OFC Engine: AI proposes, Qloo provides cultural evidence, and OFC reviews decisions before action.
-
02 — OFC ON: OFC revises Nike to Adidas using matching Qloo rankings. Commercial approval remains withheld.
-
03 — OFC OFF: Nike remains the original choice, while Qloo's verified shortlist ranks Adidas first.
Inspiration
AI agents make confident recommendations, but who verifies their decisions before they influence real-world actions?
We built OFC Engine to solve this problem.
What it does
OFC is an evidence-based decision-control layer for AI agents.
For an artist–brand partnership, the agent proposes a choice, Qloo's Taste AI provides real cultural-affinity evidence, and OFC evaluates whether to:
- Keep the original decision.
- Revise it when evidence supports a better alternative.
- Withhold it when evidence is insufficient.
How we built it
Python, Qloo Taste AI, LLMs, PostgreSQL, JavaScript, and Render.
The key innovation is our OFF vs. ON comparison: same agent, same Qloo evidence, but with or without OFC's decision controls.
Challenges and learnings
We addressed entity ambiguity, API limits, evidence verification, and durable agent workflows.
Key lesson: A convincing AI answer is not necessarily a verified decision.
Try it
Live Demo: https://ofc-fixture-preview.onrender.com/
GitHub: https://github.com/sagar-systems-lab/ofc-engine
Judge Guide: https://github.com/sagar-systems-lab/ofc-engine/blob/main/docs/JUDGE_GUIDE.md
Same AI. Same evidence. Better decisions.
Built With
- gemini
- github
- groq
- javascript
- postgresql
- pydantic
- python
- qloo-taste-ai
- render