"0xKalkiStrike team on AMD AI League Hackathon"

Unplanned machine failure costs the global manufacturing industry over $50B annually. Industrial plants capture gigabytes of high-frequency SCADA sensor streams every minute, yet operators still rely on reactive maintenance or rigid threshold alarms that trigger too late. ForgeSight AI is an end-to-end industrial intelligence platform designed to detect anomalies, predict equipment failures, explain root causes, and prescribe preventive actions before catastrophic downtime occurs. Built for AMD hardware architectures, ForgeSight AI executes a continuous 7-phase intelligence pipeline: 1. Real-Time SCADA Telemetry Ingestion: Continuously monitors vibration harmonics, thermal drift, pressure differentials, and RPM across machine fleets. 2. Machine Learning Anomaly Detection: Employs unsupervised multivariate Isolation Forest and ensemble models accelerated via AMD ROCm matrix cores for 3.8x faster inference. 3. Time-to-Failure (TTF) Risk Scoring: Calculates remaining risk horizons (e.g. 24–48 hours) and failure probabilities with contributing sensor factors. 4. Computer Vision Surface Defect Scanner: Inspects industrial parts (spindles, bearings, welds) with bounding-box defect localization (fatigue cracks, pitting, misalignment). 5. Semantic Technical SOP Search: RAG engine that queries engineering manuals to retrieve exact repair procedures, torque specs, and replacement parts. 6. Financial ROI & Downtime Avoidance: Dynamically computes saved maintenance hours and bottom-line dollar value per prevented incident. 7. Human-in-the-Loop Operator Console: Fully responsive Next.js 16 control dashboard featuring mobile emergency alert triaging and work order approval workflows. ForgeSight AI transforms factories from reactive firefighting into proactive, high-precision operations.