Pranav completed Machine Learning for Automotive Safety

Built an end-to-end perception validation dossier for a fictional Aurora-7 detector: defined detection at IoU >= 0.5, scored synthetic detections with precision/recall and VRU false-negative rate, and exposed the dangerous night and occluded slices where VRU recall collapsed. The final recommendation was evidence-based and safety-focused, using a daylight-only release condition until the top-band VRU misses are mitigated.

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