OS3: YC-Backed Robots for Real Physical Labor (2025)

OS3 is a San Francisco robotics startup founded in 2025, building semi-humanoid robots for real physical labor, backed by Y Combinator (Summer 2026 batch).

OS3 is a two-person robotics startup founded in San Francisco in 2025 by Rishabh Chanana and Chris Hailey, part of Y Combinator's Summer 2026 batch. The company designs its own robot hardware and trains its own control models together, pretraining on human video and fine-tuning with reinforcement learning on real hardware. OS3 already runs pilot deployments doing physical labor tasks in hospitality settings.

OS3 is a San Francisco robotics startup founded in 2025 by Rishabh Chanana and Chris Hailey, backed by Y Combinator's Summer 2026 batch. It builds semi-humanoid robots for real physical labor, designing its own hardware and training control models by combining human video pretraining with reinforcement learning directly on its robots. OS3 is already piloting deployments in the hospitality industry.

Founded: 2025 · HQ: San Francisco, CA, USA · Team: 1-10 · CEO: Rishabh Chanana · Funding: Y Combinator (Summer 2026 batch); additional seed investors mentioned but not publicly named or sized

About OS3

OS3 Inc. was founded in San Francisco in 2025 by Rishabh Chanana and Chris Hailey, and joined Y Combinator's Summer 2026 batch under partner Gustaf Alstromer. Chanana built his first computer at age 10, wrote his first line of code at 13, and decided a month later he wanted to start a robotics company, a thread he carried through graduate study in Robotics and AI at UC San Diego, NLP research alongside Harvard Medical School and Georgia Tech, and a stint training large language models at ServiceNow. Hailey, the company's co-founder and CTO, studied computer science at USC on a Presidential Merit scholarship, worked as a software engineer on Coinbase's asset addition team where his work touched more than $19 billion in trading volume, and turned a YC hackathon project into a venture generating $10K in monthly recurring revenue before OS3. The company is a two-person team as of mid-2026, run out of a hardware lab in San Francisco that the founders describe on their site as 'soldering in SF.' OS3 builds semi-humanoid robots designed to perform real physical labor, with a stated approach of controlling both the hardware and the intelligence layer in-house rather than buying off-the-shelf robot bodies and bolting on software. The company's site frames its philosophy directly: 'Every human starts with the same body and intelligence, then learns by living. So do our robots.' The current robot runs on 9 hours of battery life, extendable to 18, and carries two arms rated at roughly 5 kilograms of payload each on a vertical lift that spans from the ground to 6 feet, sensed with camera and LiDAR, with collision avoidance, automatic power cutoff, and an emergency stop built in. Its training pipeline has two stages: models are pretrained on internet and egocentric human video to learn general physical competence, then fine-tuned with reinforcement learning directly on OS3's own hardware, with an explicit research focus on reducing how much real-robot data that fine-tuning requires. The company has not disclosed a separate branded product name for the robot itself; OS3 is presented as one integrated hardware-plus-intelligence platform, designed and built in the US, rather than a family of named products. As of mid-2026, OS3's most concrete public milestone is an active pilot deployment in the hospitality industry, where its robots are performing physical labor tasks on real floors rather than only in a lab or simulation. The company has not published benchmark results, specific task throughput numbers, or a public roadmap of additional verticals, which is typical for a two-person, sub-one-year-old hardware startup that is still heavily in build mode. No further product launches, hardware revisions, or new pilot announcements were publicly available at the time of this listing. Funding details for OS3 have not been publicly disclosed. The company's site includes an unspecified 'Backed by' investor mention, and it is a member of Y Combinator's Summer 2026 batch, which carries YC's standard early-stage investment, but no seed round size, valuation, or named venture investors beyond YC were confirmed in public sources as of this listing. This is common for a YC company still inside or just past its batch and should not be read as an absence of investor interest, only as an absence of public disclosure. OS3 has not disclosed a revenue model publicly. Given that its only confirmed commercial activity is a hospitality pilot, the likely near-term business model is a services or lease arrangement in which OS3 places robots with a customer to perform defined physical labor tasks, similar to the deployment-first approach used by other humanoid and semi-humanoid robotics startups before they formalize pricing. No public pricing, contract structure, or customer count has been released. OS3's team consists of its two named co-founders, Rishabh Chanana and Chris Hailey, with no additional leadership, engineers, or researchers publicly named as of mid-2026. The company operates out of San Francisco with no disclosed additional offices. Given the team size, hiring is presumably a near-term priority, though OS3 has not published a public careers page or listed open roles at the time of this listing. OS3 describes itself as 'deployment-focused robotics research, across hardware and intelligence,' positioning its research agenda around closing the loop between physical hardware design and the control models that run on it, rather than treating either as a separate problem. Its technical bet, pretraining on human video before reinforcement learning on real hardware, mirrors an approach used by several other embodied-AI labs but is distinguished here by OS3 building its own robot chassis rather than adapting a third-party humanoid platform. No published research papers, safety framework, or responsible-scaling-style policy were found for the company, which is again consistent with a very early, pre-disclosure stage startup. OS3 competes in the emerging humanoid and semi-humanoid labor robotics category alongside better-funded and more established players such as Figure AI, 1X Technologies, Agility Robotics, Apptronik, and Sanctuary AI. Those competitors have raised substantially larger rounds and, in several cases, have multi-year head starts on hardware iteration and large-scale piloting. OS3's stated differentiation is tight in-house control over both the robot body and the training pipeline from day one, and an explicit focus on hospitality-sector physical labor as its first real-world proving ground, rather than warehouse or industrial logistics tasks that several rivals have targeted first. Whether that hardware-and-model co-design approach can scale against better-capitalized competitors is not yet publicly demonstrated. No EU AI Act classification, GDPR posture, export control disclosure, or other regulatory statement was found for OS3, which is expected for a two-person, pre-scale hardware company with no disclosed enterprise customers outside a single hospitality pilot. As a physical robotics company rather than a software or API vendor, standard SaaS compliance certifications such as SOC 2 or HIPAA are not applicable to its current business. OS3's near-term trajectory will likely be shaped by how its Y Combinator batch progresses, whether it discloses a funding round after Demo Day, and whether the hospitality pilot expands into a broader commercial deployment or additional verticals. As a fresh YC company with no public funding figures, product name, or headcount growth yet on record, most of OS3's story remains unwritten as of this listing and should be revisited as it discloses more.

Mission

Deployment-focused robotics research, across hardware and intelligence: robots for the real world.

Products

Links

Website · Twitter

Frequently Asked Questions

What is OS3 and what do they build?

OS3 Inc. is a San Francisco robotics startup founded in 2025 that builds semi-humanoid robots designed to perform real physical labor. Rather than buying an off-the-shelf robot body and adding software on top, OS3 designs its own hardware in its San Francisco lab and trains its control models specifically for that hardware, so the body and the intelligence layer develop together. Its training approach starts with pretraining on internet and egocentric human video so the model learns general physical movement patterns, then fine-tunes those models with reinforcement learning directly on OS3's own robots. The company has not disclosed a separate branded product name; the robot and its software are presented as one integrated platform. OS3 is already running a pilot deployment in the hospitality industry, meaning its robots are doing labor tasks on real floors rather than only in a lab. The company is part of Y Combinator's Summer 2026 batch. As a two-person team less than a year old, OS3 sits at the earliest stage of the humanoid labor robotics category, well behind larger funded players like Figure AI or 1X Technologies in scale but distinguished by its hardware-and-model co-design approach.

Who founded OS3 and who runs the company?

OS3 was co-founded by Rishabh Chanana and Chris Hailey in 2025. Chanana built his first computer at age 10 and wrote his first line of code at 13, later studying Robotics and AI in graduate school at UC San Diego, conducting NLP research alongside Harvard Medical School and Georgia Tech, and training large language models at ServiceNow before starting OS3. Hailey, OS3's co-founder and CTO, studied computer science at USC on a Presidential Merit scholarship, worked as a software engineer on Coinbase's asset addition team where his work touched more than 19 billion dollars in trading volume, and previously turned a Y Combinator hackathon project into a venture generating 10,000 dollars in monthly recurring revenue. No additional co-founders, executives, or board members are publicly disclosed as of mid-2026. The company operates as a two-person team out of San Francisco, with the founders describing their work directly as 'soldering in SF.' There is no evidence of a leadership change since founding, as the company is under one year old.

How much funding has OS3 raised?

OS3 has not publicly disclosed the size of any funding round as of mid-2026. The company's own website includes an unspecified 'backed by' investor mention without naming the investor or amount, and OS3 is a member of Y Combinator's Summer 2026 batch, which comes with Y Combinator's standard early-stage investment as part of the program. Beyond that, no seed round size, valuation, or named venture capital firm has been confirmed through public press releases, Crunchbase-style databases, or company statements. This level of non-disclosure is typical for a company still inside or just past its YC batch, which often precedes a public seed announcement around or after Demo Day. No revenue figures or ARR estimates are available, and OS3 has not indicated any IPO or acquisition plans, which would be premature for a company of its age and size.

What products does OS3 make?

OS3's core product is a semi-humanoid robot platform built for physical labor tasks, combining custom in-house hardware with control models trained specifically for that hardware. The disclosed hardware specs include 9 hours of battery life extendable to 18, two arms rated at roughly 5 kilograms of payload each, a vertical lift spanning from the ground to 6 feet, camera and LiDAR sensing, and safety features including collision avoidance, automatic power cutoff, and an emergency stop. The company has not given the robot a distinct consumer-facing product name separate from the OS3 brand itself; it is described simply as OS3's robot, an integrated hardware-plus-intelligence system designed and built in the US. The current confirmed use case is a pilot deployment in the hospitality industry, where the robots perform physical labor tasks in a live commercial setting rather than in a lab or simulated environment only. OS3 has not published a broader product roadmap, additional verticals, or a per-unit or subscription price for its robots. There is no separate developer API, consumer app, or software-only offering; OS3 is a hardware-first company where the model and machine are sold or deployed together.

Where is OS3 headquartered and how big is the team?

OS3 is headquartered in San Francisco, California, where the founders build and test their hardware in what the company describes on its own site as 'soldering in SF.' As of mid-2026, OS3's team consists of its two named co-founders, Rishabh Chanana and Chris Hailey, with no additional employees, engineers, or researchers publicly named. No secondary office locations have been disclosed. Given the team's small size and the company's age of under one year, hiring is a likely near-term priority, though OS3 has not published a public careers page or listed open roles as of this listing. The company has not disclosed any layoffs, restructuring, or major headcount changes, which would not be expected at this stage.

What is OS3's mission or research focus?

OS3 describes its work as deployment-focused robotics research across both hardware and intelligence, framed on its website with the philosophy that 'every human starts with the same body and intelligence, then learns by living, so do our robots.' Its research agenda centers on closing the loop between the physical design of a robot and the control models that run on it, treating hardware design and model training as one connected problem rather than two separate ones. Technically, its approach pretrains models on internet and egocentric human video to learn general physical competence, then fine-tunes with reinforcement learning directly on OS3's own robots, with an intended continuous learning loop so deployed robots keep improving from real operating data. No published research papers, named safety framework, or responsible-scaling-style policy have been found for OS3 as of mid-2026, consistent with a company that is still in an early, pre-disclosure stage of its research program.

Is OS3 compliant with SOC 2, GDPR, HIPAA?

OS3 has not published any compliance certifications such as SOC 2, ISO 27001, GDPR documentation, or HIPAA eligibility, and no trust center or security page was found on its website as of mid-2026. This is expected given the company's stage: OS3 is a two-person hardware and robotics startup with a single disclosed hospitality pilot rather than an API or SaaS product handling third-party customer data at scale, so the compliance certifications typical of a software vendor are not yet applicable to its business. No EU AI Act classification, data residency commitment, or export control disclosure has been made public. As OS3 scales into more commercial deployments, formal compliance documentation would be expected to follow, particularly if it begins serving enterprise hospitality customers with data handling requirements.

Who are OS3's main competitors?

OS3 competes in the emerging humanoid and semi-humanoid labor robotics category alongside better-funded, more established companies including Figure AI, 1X Technologies, Agility Robotics, Apptronik, and Sanctuary AI. Those competitors have raised substantially larger disclosed funding rounds and, in several cases, have multi-year head starts on hardware iteration, safety testing, and large-scale piloting across warehouse and industrial logistics settings. OS3's stated differentiation is designing its own robot hardware and training pipeline together from day one rather than adapting a third-party humanoid platform, paired with an early focus on hospitality-sector physical labor as its first real-world proving ground instead of warehouse logistics. OS3 currently has no disclosed answer to the large-scale industrial deployments that rivals like Agility Robotics have already demonstrated, and its unfunded, two-person scale means it cannot yet match competitors on iteration speed or unit volume. As a brand-new Y Combinator company, OS3 itself is an emerging name that established robotics investors were not tracking closely before its 2026 batch.