AI / Predictive Maintenance · 2026
Lumin.AI
Track Winner & Top-5 National at HACKaMINeD 2026 — AI solar maintenance.
Lumin.AI is an AI-powered predictive maintenance platform for solar inverters. A 7-stage ETL pipeline ingests inverter telemetry, which feeds a hybrid risk engine combining Isolation Forest for unsupervised anomaly detection with XGBoost for supervised fault classification. Predictions are served as a FastAPI microservice with SHAP-based explainability, enabling field engineers to understand why a unit is flagged at risk. The project won the Track at HACKaMINeD 2026 and placed Top-5 nationally across 2,200+ participants.
Specifications
Features
7-stage ETL pipeline for solar inverter sensor data ingestion and normalisation
Hybrid Isolation Forest + XGBoost ensemble risk scoring engine
SHAP feature attribution for interpretable, auditable risk predictions
FastAPI microservice with real-time inference endpoint
Track Winner & Top-5 National at HACKaMINeD 2026 (2,200+ participants)