Nandigam Harsha Vardhan completed Machine Learning for Automotive Safety

Built a perception validation dossier that correctly defined detection at IoU >= 0.5, scored a synthetic detections table, and separated aggregate performance from the safety-critical VRU false-negative rate. The work then drilled into day/night, occlusion, and rarity slices to surface the highest-risk misses and paired them with concrete SOTIF mitigations such as sensor-fusion fallback, reduced-speed operation, and targeted data expansion.

• Leakage-Free Detection Definition
• Safety-Critical Metric Selection
• Per-Condition Failure Analysis
• SOTIF Functional-Insufficiency Reasoning