Ideas made tangible
A selection of systems that turn ambiguous requirements into useful, measurable products.
Problem: coaching feedback is generic and slow. What I built: a Flask + Azure OpenAI platform with a FAISS RAG engine aligned to GROW and STAR frameworks, plus sentiment analytics that surface real-time feedback across 5+ communication metrics. Result: 95% framework adherence on coaching responses.
Problem: transformer waste heat is lost as unusable energy. What I built: a thermoelectric system converting transformer waste heat into usable DC electricity, with a real-time CMS streaming sensor data. Result: a published patent application (202541114252 A) with predictive maintenance and a ROS2 digital twin in development.
Problem: business-critical asset records were hard to track and audit. What I built: a Django + PostgreSQL lifecycle platform with a complex relational schema, enterprise IT compliance and RBAC. Result: managed 1,000+ business-critical asset records across production operations.
Problem: field technicians had no reliable live status for service tickets. What I built: NestJS APIs and MySQL service-ticket operations, with Firebase Cloud Messaging pushing updates to Android and cross-platform technicians. Result: real-time ticket visibility for 1,000+ records.
Every project is designed to move a real-world signal toward a useful outcome. Select a node to inspect its role.
A real operational or human problem starts the flow.
A retrieval-augmented generation chain: the question is guarded, grounded in evidence, answered by a hosted model and evaluated before it ships.
A real question enters the system — from a chat window, API client or workflow. The raw string is the only signal that starts the chain.