At CES 2026 in January, a KUKA industrial arm and an ASUS edge box shared a booth demo with a small Swiss company most of the audience had never heard of. The arm slowed, paused, and resumed as humans stepped in and out of its work envelope. No cameras. No lidar. Just an Ultra-Wideband radar chip of the sort that already ships in millions of cars and phones, running software from a Lausanne-area startup called Algorized [Algorized].
That demo, alongside a quieter one with KUKA, is the elevator pitch for what Algorized just raised $13 million in Series A funding to scale [Algorized]. The company calls its product an "edge-native nervous system for physical AI," which is a mouthful, but the underlying bet is refreshingly concrete: the wireless radios already embedded in cars, robots, laptops, and home appliances can be taught, through software alone, to detect human presence, track position, and even read vital signs through walls and obstacles [Algorized]. No new silicon required.
The bet
Algorized sells a foundation model and SDK for people-sensing on commodity wireless sensors, principally UWB, mmWave, and Wi-Fi [Algorized]. The wedge is a software-only upgrade path. A semiconductor partner like Qorvo, whose UWB Partner Program Algorized joined in 2024, can offer customers a richer feature set without taping out a new chip [Qorvo]. An automotive Tier 1 can add child-presence detection to a model year already locked into specific radar hardware. That last use case is not hypothetical: in March 2026, Algorized announced a collaboration with ARIA Sensing to deliver AI-powered UWB radar for next-generation automotive child presence detection [Aria Sensing, 2026], a feature that European safety regulators are pushing toward standard.
The customer, then, is not the end consumer. It is the chip maker, the Tier 1, the robot OEM, and the AIoT integrator. ASUS IoT signed a strategic partnership with Algorized in 2025 to combine the model with ASUS edge hardware for AIoT deployments [ASUS Pressroom, 2025]. The company says it is now deployed with major automotive and industrial leaders in Europe and the U.S. [Algorized], and LinkedIn material describes deployments across retail, automotive, robotics, and semiconductors [LinkedIn].
Why it could be big
The physical-AI thesis, that robots and vehicles need a cheap, privacy-respecting way to know where humans are, has attracted serious capital. Amazon's Industrial Innovation Fund led Algorized's $4.3 million seed [Algorized], a notable signal given Amazon's own warehouse robotics fleet and the safety-cage problem it has been chipping away at for a decade. The Series A added $13 million [Venturelab], with Acrobator, the University of California, 432 Legacy, and Run Ventures on the cap table. MassRobotics in Boston brought Algorized into its accelerator to plug it directly into U.S. robotics OEMs [Algorized].
Seed (Amazon IIF lead) | 4.3 | $M
Series A | 13.0 | $M
Total disclosed | 17.3 | $M
Back of envelope on the addressable opportunity. Roughly 85 million new cars are sold globally each year. If even a third end up with UWB anchors for digital key and child-presence detection by 2028, that is around 28 million vehicles annually. License a people-sensing software stack at, say, $2 to $5 per vehicle (estimated, in line with typical Tier 1 software royalties), and the automotive slice alone is a $55 to $140 million annual royalty pool. Add industrial robotics, where a single collaborative cell that avoids one OSHA-reportable incident pays for a lot of software licenses, and the unit economics start to look like the kind of high-margin IP licensing business that European deep tech occasionally produces and rarely scales.
The team and traction
CEO Natalya Lopareva founded Algorized after running ARHUB, an AI-enabled technologies company, from 2016 to 2023 [Clay]. She led the Swiss National Startup Team at MWC25 in Barcelona [Algorized], which is the kind of soft signal Swiss innovation agencies do not hand out casually. The company is hiring software engineers and a software architect in the U.S. [Algorized], consistent with the Series A's stated aim of expanding North American deployments. The technology itself was founded on extensive academic research [Algorized], a polite way of saying the signal-processing heritage runs through Swiss labs that have been doing radar work for a long time.
The honest counterfactual
The bear case is straightforward. Algorized is a software layer that depends on someone else's silicon shipping in volume, with the customer's BOM cost and design cycle determining whether the model ever gets loaded. UWB adoption in automotive has been steady but not explosive, and competing approaches (in-cabin camera, capacitive seat sensors, 60 GHz radar from incumbents like Infineon and Texas Instruments with their own software stacks) are fighting for the same socket. The bull answer, supported by the Qorvo, ASUS, and ARIA partnerships [Qorvo] [ASUS Pressroom, 2025] [Aria Sensing, 2026], is that Algorized has chosen to be the Switzerland of people-sensing software, neutral across silicon vendors, which is exactly what a Tier 1 wants when it does not know which radar chip will win the decade. Being the model that runs on all of them is a defensible position if execution holds.
What to watch
The next twelve months should answer two questions. First, does the KUKA and ASUS CES 2026 demo convert into a named, paying industrial deployment with disclosed volumes? Second, does at least one automotive program-win get announced ahead of the EU's tightening child-presence detection requirements? A Series B in late 2026 or early 2027 would be the natural follow-on if either lands. Watch the U.S. engineering hires close: a software architect req in the U.S. usually precedes a Detroit or Silicon Valley customer, not a press release.
The company Algorized most needs to beat is not another startup. It is Infineon, whose XENSIV radar family ships with its own people-sensing reference software and has the incumbent advantage with every Tier 1 already on speed dial. Algorized's wager is that a silicon-neutral foundation model will outrun a silicon vendor's captive stack. That is a bet worth watching.