Ingres AI

The Problem

Groundwater data for India exists, but it lives in technical datasets that researchers and local officials can't query quickly. Getting a simple answer about water levels in a region means digging through complex geological data.

What I Built

A microservices platform with a Go backend and a React frontend. An HTTP backend manages users and sessions, and an agent service handles tool calling against the groundwater data. LLM providers are pluggable: switching between Groq and Gemini is one environment variable, and a factory handles the API differences. Goroutines and channels keep the tool-calling loop and response generation concurrent and thread-safe, and a Redis cache cuts latency for frequent queries.

Related Projects