@permanatedy927

TokenPilot is an intelligent, hybrid AI agent designed for enterprises that want to control their AI spend without sacrificing accuracy or user experience. Built for AMD Hackathon Track 1, it acts as a smart router that evaluates incoming natural language tasks across 8 different categories and dynamically dispatches them to the most cost-effective solver. To minimize external API calls, TokenPilot uses a three-tier architecture: Deterministic Engine: Mathematical reasoning problems (such as arithmetic, compound interest, and rates) are parsed using robust regex and computed directly in Python, consuming exactly zero tokens with 100% precision. Local Edge Inference: Tasks like sentiment analysis, named entity recognition, summarization, and basic factual/logical questions are safely routed to a local Ollama instance running Qwen 2.5 3B. This allows the system to process the majority of typical workloads locally at zero cost. Premium API Fallback: For highly complex tasks—such as code generation or advanced logical deductive reasoning—the system escalates to Fireworks AI's premium models (kimi-k2p7-code and minimax-m3) to ensure the accuracy gate is cleared. By intelligently analyzing task difficulty and utilizing local compute first, TokenPilot successfully processes up to 77% of general-purpose tasks without a single API call, demonstrating a highly efficient, scalable, and cost-effective approach to enterprise AI routing.

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.