Self-directed project
Machine Learning for Automotive Safety
Perception Validation Engineer (ADAS Safety Sign-off) · AI / ML · August 2026
A ProoV case study · educational project, not employment
Built a perception validation dossier that defined detection by IoU threshold, scored a synthetic detections table, and separated aggregate performance from the safety-critical VRU false-negative rate. The analysis then broke recall down by lighting, visibility, and rarity, correctly exposing the night and occluded slices as the highest-risk failure modes.
Graded against
- Detection metrics correctness (precision, recall, VRU false-negative rate)30%
- Failure-mode analysis & per-condition rigour25%
- SOTIF safety case quality (ISO 21448 reasoning)25%
- Ship verdict — evidence-based & internally consistent20%
Passed · pass mark 60/100
What stood out6
- Correctly locked the IoU threshold rule and showed a worked box-overlap example with a passing match at IoU 0.806.
- Computed aggregate precision 0.95, recall 0.87, and the VRU false-negative rate 0.24 from the scored table.
- Sliced VRU recall by condition and surfaced the dangerous night/occluded collapse, with worst-slice recall at 0.27.
- Leakage-Free Detection Definition
- Safety-Critical Metric Selection
- Per-Condition Failure Analysis
The work I submitted10 tasks
5 tasks · 8.9k characters · Python · 5 written answers · 721 words
- Example Iou Demo1.5k characters
- Example Metrics DemoPython · 2.2k characters
- Scored Metrics RunPython · 2.3k characters
- Ttcs Example SlicePython · 1.3k characters
- Ttcs Triage SlicePython · 1.6k characters
- Example Iou Demo169 words
- Example Metrics Demo154 words
- Scored Metrics Run143 words
- Ttcs Example Slice143 words
- Ttcs Triage Slice112 words
Summarised on purpose — the submitted code and writing stay private so this page cannot be reused as an answer key. The full submission sits behind the verified certificate.