Muhammad Muneeb@mmuneebrehman394

Traditional AI agents often waste expensive cloud API tokens on simple tasks or hidden "reasoning" generation, resulting in massive operational costs. Our solution introduces a sophisticated 5-layer fallback pipeline designed to maximize task accuracy while keeping inference costs near zero whenever possible. Built to survive strict grading constraints (4 GB RAM, 2 vCPU, 10 GB Docker limit), the agent natively resolves straightforward queries using deterministic solvers and a bundled local LLM (Qwen 2.5 3B). Only when tasks exceed local capabilities does the system dynamically route requests to the cloud. When utilizing the Fireworks API, it selects the most cost-effective model and actively suppresses hidden reasoning tokens to conserve budget. Paired with a bounded retry mechanism to guarantee JSON schema compliance, this architecture is a cost-aware, fault-tolerant AI agent tailored for strictly bounded environments.

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.