Atharva Pardeshi completed Machine Learning for Automotive Safety

Built an end-to-end perception validation dossier for a fictional Aurora-7 detector: defined detection using IoU >= 0.5, scored synthetic detections with precision/recall plus VRU false-negative rate, and sliced VRU recall across day/night, occlusion, and rarity to expose the highest-risk misses. The final recommendation was evidence-based and conditional, tying the release decision to the detector’s weakest safety slices and concrete mitigations.

• Leakage-Free Detection Definition
• Safety-Critical Metric Selection
• Per-Condition Failure Triage
• Evidence-Based Release Recommendation