About NADIR
NADIR builds forensic reconstruction for crashes involving driver-assistance systems: what the car could perceive, and whether it should have acted.
We started out building real-time ADAS sensor-drift detection from fleet telematics. An internal review found that the telematics signal cannot actually detect camera drift and that all of our evidence was synthetic. We stopped selling it, archived it, and kept the parts that transfer: a dashcam perception stack, SE(3) and factor-graph estimation code, changepoint detection, evidence hashing and signing, and a method for studying a production ADAS only through what it exports. The forensics product is built on those, held to a stricter rule: synthetic data is for debugging, and the demo never gets ahead of the benchmark.
Founders
- Dhruv Hegde
- CEO. Leads the reconstruction engine and technology. Computer Science and Mathematics, University of Michigan.
- Sri Balaji
- COO. Leads go-to-market and operations, including work with reconstruction experts and test facilities. Computer Engineering, University of Michigan.
NADIR has no other staff or advisors at present.
NADIR Research
NADIR Research is the division that defines how our methods are measured and publishes the results as numbered technical reports, with DOIs, before the product relies on them. Its first forensic report is the benchmark protocol.