Congruent

We build radars for end-to-end autonomy

Winter 2026ActiveIndustrialsRoboticsRadarAIAutomotive

Company

At Congruent, we build radars for end-to-end autonomous systems. The most advanced autonomous systems are trained as a single neural network from raw sensor data to navigation actions. For a sensor to be included in these pipelines sensor stacks requires two key properties: access to raw sensor data and a high-fidelity sensor simulator. Current automotive radars have neither, they output heavily processed point clouds and no raw radar simulator exists for driving scenes. Congruent solves both problems: a radar architecture that exposes raw data, paired with a world model based radar simulator. Radar is the only depth sensor at a price point that scales to every car on the road and works in all weather conditions. Congruent is building the radar compatible with the training architectures that will make mass-market vehicles autonomous.

Verdict

High Signal

Market Opportunity

Autonomous vehicle sensor market is multi-billion dollar with clear B2B ICP: AV developers, Tier 1 automotive suppliers, and robotics companies needing radar compatible with end-to-end training pipelines. Radar is uniquely positioned as the only all-weather depth sensor at price points that scale to mass-market vehicles, making the TAM essentially the entire automotive autonomy stack.

High Signal

Founder Signal

Clement Barthes: PhD UC Berkeley Structural Engineering, 5+ years as Lab Manager at Berkeley PEER, 5+ years as CTO at Safehub (sensor startup), then 2.75 years as ML Manager at Zendar (radar-specific AV company). Evan Carnahan: PhD UT Austin Geophysics, 1.5 years at NASA JPL, then 3.5 years at Zendar progressing from intern to Research Engineering Manager leading radar ML/perception teams. Both founders have deep, directly relevant radar + ML + autonomous systems experience from exactly the right prior company (Zendar).

Medium Signal

Competition

Traditional automotive radar vendors (Bosch, Continental, Aptiv, ZF) output processed point clouds incompatible with end-to-end training, not raw data. Radar simulation startups like Metawave or established players don't appear to offer the raw-data + digital twin combination Congruent claims. However, well-resourced competitors like Wayve, Tesla, or large Tier 1s could develop similar capabilities internally, and the hardware moat is unproven at this stage.

Low Signal

Product

Website shows a physical radar hardware product with a digital twin simulator concept, but no named customers, no pricing, no API docs, no revenue or usage metrics. Only a 'Book a Demo' CTA and YC backing badge. Hardware prototype appears to exist given the product render, but no evidence of deployment or paying customers.

OverallB Tier

Two deeply qualified co-founders who both came directly from Zendar — the exact radar-for-AV company — gives this team exceptional domain credibility that's rare at the YC stage. The technical thesis is sharp: current radars output processed point clouds incompatible with end-to-end training, and no raw radar simulator exists. However, this is still early-stage hardware with no visible customers, revenue, or deployed units, and hardware startups carry enormous execution risk around manufacturing, cost, and sales cycles. The market timing is right as end-to-end autonomy (Tesla FSD, Wayve) becomes dominant, but Congruent needs to show paying customers soon to validate that AV developers will actually buy external radar hardware rather than build in-house.

Active Founders

Clement Barthes

Co-Founder

ex-ML engineer and manager at Zendar ex-CTO at Safehub, making smart sensors to evaluate building damage after earthquakes ex-Research Engineer and Lab Manager at UC Berkeley - PEER lab

Evan Carnahan

Co-Founder

Co-Founder @ Congruent | Machine learning researcher with a deep background in signal processing and sensor fusion. Compulsive generalist and deeply curious about all things sensing and learning.

Congruent

TierB Tier

BatchWinter 2026

Team Size2

StatusActive