Muhammad Wali completed Machine Learning for Automotive Safety

Built a perception-validation analysis that correctly defined detection matching with IoU, computed aggregate precision/recall plus the safety-critical VRU false-negative rate, and then stress-tested performance by condition. The work surfaced a sharp night-and-occlusion recall collapse and translated it into a clear risk signal for vulnerable road users.

• Leakage-Free Detection Definition
• Honest Metric Evaluation
• Per-Condition Failure Analysis
• Safety-Critical Metric Prioritization