Adarsh S completed Machine Learning for Automotive Safety

Built an end-to-end perception validation dossier for a synthetic Aurora-7 detector: defined detection at IoU ≥ 0.5, scored aggregate and VRU-specific precision/recall, and exposed the safety gap by slicing recall across night, occlusion, and rare-class conditions. The final release recommendation was tied directly to those measured weaknesses and to SOTIF-style mitigations rather than generic accuracy claims.

• Leakage-Free Detection Metric Evaluation
• Condition-Based Safety Analysis
• SOTIF Functional-Insufficiency Reasoning
• Evidence-Linked Release Recommendation