alex says completed Machine Learning for Automotive Safety

Built a perception validation dossier that correctly defined detection via IoU, computed precision/recall and VRU false-negative rate from a scored synthetic table, and then stress-tested VRU recall across day/night, clear/occluded, and common/rare slices. The work shows disciplined safety analysis by prioritizing vulnerable road user misses over headline accuracy and by isolating the hardest operating conditions where recall collapses.

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