MRRPhysical AI
intelligence
← Company intelligence

Physical AI software · early

HYPRLABS

Corporate marketdevtoolsCurated record
hypr.co

AI robotics company building HYPRDRIVE / HYPR ACTV, a 'no-priors' robot autonomy architecture based on Run-time Learning (RTL) that rejects hand-coded rules, HD maps and classical simulation in favor of continual learning directly from real-world interaction. Demonstrated a 20-minute San Francisco drive using five cameras and a single NVIDIA Orin AGX at ~33W, and plans to integrate the stack into its own next-generation robots with a first product in 2026. Co-founded by ex-Zoox founding CEO Tim Kentley-Klay and MuZero implementer Werner Duvaud.

HeadquartersSan Francisco, US
Founded2024
Corporate funding$21M
TeamNot listed
Deployments0
Open roles6
01

Company overview

Identity and operating footprint

Company typePhysical AI software
Market segmentCorporate
StageVenture (stealth emergence Dec 2025)
FoundersTim Kentley-Klay (Co-founder & CEO), Werner Duvaud (Co-founder)
Regions servedNorth America, Europe
Countries deployedNot listed
Service footprintNot listed
Last reviewed2026-07-05
LinkedInNot listed
CA

Funding history

Publicly announced rounds

Investors on recordPrincipal Venture PartnersBlackbird
HR

Hiring signal

Public job board · 2026-07-13

Open roles6

Hiring is an operating signal, not deployment evidence. Roles are refreshed from the company’s public careers system.

TL

Company timeline

Milestones and announcements

2025-12-15Emerged from stealth with HYPRDRIVE, a run-time-learning autonomy architecture; demoed a 20-min SF drive on 5 cameras and a single Orin AGXPR Newswire
2025-12-15HYPRLABS and the case for 'no-priors' physical AI ↗Principal Venture Partners
SC

Source ledger

3 unique public sources

01
HYPRLABS Emerges from Stealth with HYPRDRIVE, A No-Priors AI ArchitectureYahoo Finance / PR Newswire · retrieved 2026-07-05
02
SEC Form D: $6,258,500 sold of $6,258,500 offeringSEC EDGAR · retrieved 2026-07-05
03
HYPRLABS Emerges from Stealth with HYPRDRIVE, A 'No-Priors' AI Architecture Learning Directly from RealityPR Newswire · retrieved 2026-07-05