Zhenyu Jiang@Jamix

1. Overview & Value EcoRoute is a production-grade, multi-model hybrid orchestration agent developed for the AMD Hackathon (Track 1). It serves as an automated cost-optimizer layer that sits between users and diverse backends, routing intelligence seamlessly without modifying core prompt intents. By dynamically matching query complexity with the most cost-effective model under strict resource limits, EcoRoute bridges the gap between budget constraints and high-quality outputs. 2. Dual-Track Token Crushing Strategy Built to dominate on token efficiency under strict accuracy gates, EcoRoute employs a custom Dual-Track Adaptive Routing Strategy: Zero-Token Local Track: For structural text processing and low-complexity extraction tasks—specifically Sentiment (Cat 3), Summarisation (Cat 4), and NER (Cat 5)—the container invokes a local, 4-bit quantized Gemma-4-E2B model via llama-cpp-python. Operating purely within the grader's 4GB RAM / 2 vCPU limits, these local completions drain exactly zero Fireworks tokens, securing a massive advantage on the leaderboard. High-Fidelity Remote Track: For computationally demanding reasoning domains—Math (Cat 2), Debugging (Cat 6), Logic Puzzles (Cat 7), and Code Gen (Cat 8)—the agent cleanly dispatches requests to dedicated online experts (such as kimi-k2p7-code or gemma-4-31b-it) to pass strict accuracy thresholds. 3. Industrial-Grade Fault Tolerance Built with an advanced Catch-All Exception Shield, EcoRoute offers unmatched resiliency. If the local Gemma-4-E2B runtime encounters any hardware-level virtualization or instruction set limitations, the exception shield seamlessly hot-swaps the affected tasks to the cheapest available online API node in real-time. This architecture guarantees a successful Exit Code 0 under any extreme edge case, protecting the evaluation pipeline from unexpected failure states.

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.