Beneyas Tadu@thedigitalbeni

HybridRoute is a token-efficient routing agent built for the 8 published capability categories: factual knowledge, math reasoning, sentiment classification, summarization, named entity recognition, code debugging, logical reasoning, and code generation. The core idea: local inference costs zero tokens on the leaderboard, so a quantized Qwen2.5-3B model bundled directly in the Docker image handles sentiment, summarization, and named entity recognition, all categories verified correct through repeated real-prompt testing. Categories needing precise multi-step reasoning or syntactic correctness (factual knowledge, math, logic, code debugging, code generation) route to Fireworks, using MiniMax M3 for reasoning-heavy tasks and Kimi K2.7 Code for programming tasks, chosen per category through direct side-by-side accuracy and token-cost comparison rather than assumption. The agent includes a fallback chain if a preferred model fails or returns an empty response, and every remote call is capped with a category-specific token budget tuned from real failure cases encountered during testing, including a bug where models silently truncated answers while reasoning internally despite explicit instructions not to. The full pipeline was validated end-to-end under the actual 4GB RAM / 2 vCPU grading constraints before submission, with repeated clean runs confirming stability.

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.