Mo Ashaibani@mashaibani

yassai is an AI agent built on two principles: the simplest action space that is still fully capable, and the fewest tokens spent to reach a correct answer. A small custom classifier runs in-process and tags every task with its capability categories, enabling per-category routing, reasoning calibration, and prompt tailoring. Two fine-tuned local models do the work: a MiniCPM5-1B tool lane that solves maths and logic through executed code, and a Qwen3.5-2B assist lane for everything else. Deterministic evidence gates verify each answer before it is trusted, and only gate-rejected tasks fall through to a single remote insurance call, so tokens are spent on answers rather than orchestration overhead.

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.