decision-agent

The product accepts a company name, runs a multi-step research workflow, and produces an investment decision with supporting evidence. The workflow is designed to be explainable, modular, and easy to extend. Each analysis stage is isolated into its own agent so the system can evolve from a prototype into a production research platform

Edge AI PoC simulating a smart tomato greenhouse: CNN-based leaf disease detection, compressed for Raspberry Pi, driving a rule-based agent and real-time Streamlit dashboard.

Native analytics app for post-move demand forecasting, district-level action cards, and operations monitoring.