Gurkeerat Kang 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, computed precision/recall and the VRU false-negative rate from a scored table, and exposed the dangerous night and occluded slices where VRU recall collapsed. The final recommendation was evidence-based and conditional, with SOTIF-style mitigations tied directly to the highest-risk failure modes and a correct comparison to the Uber Tempe crash.

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