FieldSpace
ABOUT THE COMPANY

Safety case evidence that replays the same way twice.

We're a small, senior team building the deterministic evidence layer for autonomous fleet safety cases. FieldSpace populates the behavioral acceptance criteria in the industry-standard safety case framework with per-event evidence that operators, auditors, and underwriters can re-run identically.

What we believe

Sell to the entities carrying the safety liability

AV operators publish safety cases. Insurers underwrite the resulting fleets. Both are budget-owners with a real problem. Tier-1 suppliers offload liability through contracts, which is why a third-party evidence layer was never their budget line.

Determinism is useful because it is reviewable

A safety team, an auditor, and an underwriter should all be able to re-run the same evidence and get the same output. Sampled neural runs cannot. Exact-number-theory determinism can.

Map into the framework that already exists

The Favaro 2023 paper invited third-party evaluative methodologies to populate the behavioral acceptance criteria space. FieldSpace is one of those methodologies, not a replacement for the operator's existing portfolio.

Source review should be available when it matters

For serious safety review, customers need more than a demo. Under NDA, the safety path should be inspectable enough for engineers to challenge assumptions and reproduce results.

Where we came from

FieldSpace started as an internal research programme inside StarDrive, a deep-tech company building physics-grounded infrastructure for space situational awareness and defence autonomy. The same mathematical primitives (partial differential equations solved at wire-speed on commodity silicon) that let us propagate satellite uncertainty turn out to work remarkably well for propagating traffic risk in a driving scene.

We took that kernel and wrapped it in a safety observer, benchmark harnesses, and a closed-loop planner wrapper called FieldSpace. Today the focus is public benchmark evidence, log replay, and narrow partner-selected edge cases.

The team is senior, small, and technical. We've built satellites, fighter-jet avionics, and production driving stacks at companies whose names we can't put on a marketing page. We'd rather ship than brand.

The facts that fit on one page

Founded
2024
Stage
Pre-seed / seed
HQ
San Francisco
Industry
Automotive · Defence · Robotics
Parent
StarDrive, Inc.
Program
NVIDIA Inception

Short calls. Bring the safety case.

If you're authoring an AV safety case, underwriting an autonomous fleet, or running a safety-case program inside an OEM or Tier-1, we'd like 30 minutes.