An AI-powered Advanced Rider Assistance System (ARAS) that uses computer vision to detect road threats in real time and warn motorcycle riders before a collision happens. I worked on improving model accuracy, optimizing resource usage, and helping migrate the solution to a new edge AI platform.

Enhanced front‑collision and safety monitoring using optimized ML models.

Reduced inference resource usage on edge devices while preserving real‑time performance.

Migrated the computer‑vision pipeline from Jetson to a Hailo‑based edge AI platform.

Shipped and validated the system with thousands of riders in real‑world traffic conditions.

Frontend: Edge device interface and real-time visualization components for rider alerts and system status.

Backend: Python-based ML pipeline with XGBoost models, C++ for performance-critical components, ONNX runtime for optimized inference, and real-time data processing pipelines.

Services & Infrastructure Hailo edge AI platform, cloud analytics services for fleet data aggregation, and real-time streaming infrastructure for safety event processing.

Other Technologies Computer vision algorithms, model optimization and pruning techniques, edge computing architecture, and IoT device integration.