Building trust in digital evidence
Axiom is developing physics-based verification for video evidence, starting with insurance claims workflows where capture integrity directly impacts automation and trust.

Adhiraj Chhoda
Founder & Researcher
I started Axiom after my aunt lost her savings to a deepfake video call. Watching someone I love get deceived by synthetic media that looked completely real made the problem personal, and made me realize that detection after the fact isn't enough.
The insight that led to Axiom: instead of trying to detect fakes, verify authenticity at the point of capture using physics that can't be faked. Sensor data from the device, matched against what the video shows, creates constraints that synthetic content can't satisfy.
I'm building Axiom to make verification infrastructure that enterprises can trust, starting with insurance claims, where the cost of unverified evidence is measured in fraud losses and manual review hours.
Why Axiom
The problem with "detecting" fakes
Most approaches to synthetic media try to detect artifacts: compression patterns, facial inconsistencies, audio mismatches. But this is fundamentally a losing game. As generation improves, artifacts disappear. Detection becomes a constant arms race.
Axiom takes a different approach: instead of asking "is this fake?", we ask "was this captured by a real camera in a real physical scene?" The answer comes from physics: synchronized sensor data that must match what the video shows, in ways that synthetic content cannot replicate.
This only works when you control capture. That's why we're starting with professional workflows, like insurance adjusters and field inspectors, where requiring our capture module is operationally feasible and the ROI is immediate.
Our Approach
Physics-constrained verification
Axiom captures video alongside synchronized IMU telemetry. Our verification pipeline cross-validates optical flow against sensor-predicted motion, using rolling-shutter timing as a dense temporal constraint. The output is a three-state decision: verified, violation detected, or inconclusive. This routes workflow without making accusations.
Read the technical specificationGet in touch
Interested in learning more about Axiom, exploring a partnership, or have questions about our approach?