Sagacious
Team consisting of an Assembled algorithmic engineer (PhD, Operations Research) and a Google systems engineer, specializing in mathematical optimization, Python, C++, and AI agents.
Project Description
Our core innovation is a multi-stage intelligent routing engine that moves beyond simple confidence thresholds. We implemented a “common sense” heuristic that evaluates message size and input complexity before execution; short, repetitive tasks are handled locally to preserve privacy and speed, while massive context windows or ambiguous queries are dynamically escalated to Gemini 2.0. To empower the cloud model, we enriched the system prompt with few-shot examples and task-specific context, significantly improving tool-call correctness. This architecture ensures a seamless user experience while maximizing local efficiency.