Aryan Sharma completed Machine Learning for Automotive Safety

Built a complete perception-validation dossier for a synthetic Aurora-7 detector: defined detection correctly with an IoU >= 0.5 match rule, scored the table to obtain aggregate precision/recall and a VRU false-negative rate, then sliced recall by lighting and occlusion to surface the night and occluded pedestrian risks. The final recommendation was evidence-based and conditionally scoped, with SOTIF-style mitigations and a real-case comparison to the Tempe crash.

• Leakage-Free Detection Definition
• Honest Safety Metric Evaluation
• Per-Condition Failure Triage
• SOTIF Functional-Insufficiency Reasoning