Eco Router AI

Created by team Team Skem on July 06, 2026
Hybrid Token-Efficient Routing Agent

Eco-Router-AI is a general-purpose AI agent built for the hackathon’s Track 1 challenge. It takes natural-language tasks from JSON input, understands what kind of request each prompt represents, and routes it to the most suitable allowed Fireworks model at runtime. The goal is to deliver strong answers while keeping inference efficient, so the system is designed around smart model selection instead of using the same model for every request. It supports a broad range of task types, including reasoning, summarization, sentiment analysis, named entity recognition, code debugging, and code generation, which makes it useful for mixed and unpredictable workloads. The pipeline is also built to produce clean JSON output, track analytics, and operate within the competition’s runtime and token-efficiency constraints.

Category tags: