VYSAKH M completed Machine Learning for Automotive Safety

Built a perception validation dossier that defined detection with an IoU ≥ 0.5 threshold, scored a synthetic detections table, and separated headline metrics from the safety-critical VRU false-negative rate. The dossier then sliced recall by night, occlusion, and rarity, triaged the highest-risk misses, wrote ISO 21448 SOTIF chains with concrete mitigations, and issued an evidence-cited conditional ship recommendation tied to the Uber ATG Tempe crash.

• Leakage-Free Safety Metric Framing
• Condition-Sliced Failure Analysis
• SOTIF Functional-Insufficiency Reasoning
• Evidence-Based Deployment Verdicting