Skip to content
HOMETECH DIRECTORYNEWS & DEVELOPMENTSRELOCATEEMPLOYERSNEIGHBORHOODSABOUTSchedule a Call

Voxel51

Voxel51 is the Ann Arbor AI infrastructure company behind FiftyOne. Founded 2016 by U-M's Jason Corso and Brian Moore. Bessemer led a $30M Series B in M...

A2Hub Editorial· Rusty Roof Media· Updated May 4, 2026

Voxel51

Voxel51 is the clearest example in 2026 of what Ann Arbor applied AI looks like. A University of Michigan spinout building developer infrastructure that sits underneath a large share of serious computer vision work. Its open source project FiftyOne is a standard tool inside enterprise ML teams, and the company closed a $30M Series B led by Bessemer Venture Partners in May 2024.

Quick facts

  • Founded: 2016
  • Founders: Jason Corso (Chief Scientist), Brian Moore (CEO)
  • HQ: Ann Arbor, Michigan
  • Headcount: Growth-stage; specific 2026 headcount not publicly disclosed
  • Funding: Approximately $45.5M total raised
  • Rounds: $1.25M federal grant, $2M seed, $12.5M Series A (2022, Drive Capital and Shasta Ventures), $30M Series B (May 2024, Bessemer Venture Partners lead)
  • Key investors: Bessemer Venture Partners, Drive Capital, Shasta Ventures, Tru Arrow Partners, ID Ventures
  • Current leader: Brian Moore, CEO
  • Sector tags: AI, computer vision, MLOps, developer infrastructure
  • Careers: voxel51.com/careers
  • Website: voxel51.com
  • LinkedIn: linkedin.com/company/voxel51

What Voxel51 makes

Voxel51 builds developer infrastructure for computer vision.

The flagship product is FiftyOne, an open-source Python toolkit for building high-quality computer vision datasets and evaluating the models trained on them. FiftyOne lets a machine learning engineer load a dataset of millions of images or video frames, explore them visually, surface mislabels and edge cases, run queries against embeddings, and iterate on data quality with the same rigor software engineers apply to code.

The thesis is that the hard problem in computer vision is not the model, it is the dataset. Teams were training on dirty, unlabeled, redundant, skewed image and video data and getting the results you would expect. FiftyOne is positioned as the tool that fixes that, either in isolation or as the data layer inside a broader MLOps stack.

On top of the open-source core, Voxel51 sells FiftyOne Teams, an enterprise tier that adds collaboration, access controls, scalable storage backends, and integrations with the MLOps tools companies already run. The product integrates with annotation platforms, experiment trackers, cloud object stores, and model training frameworks. Voxel51 has publicly announced partnerships with Databricks (working with FiftyOne data directly inside Databricks environments) and with V7, whose Darwin annotation platform is commonly used by computer vision teams.

Customers span autonomous driving, robotics, medical imaging, defense, retail analytics, and industrial inspection. The positioning is developer-first. FiftyOne is a pip install. Documentation, tutorials, and community presence are aimed at machine learning engineers who live in Python notebooks. That posture is what gave the company organic bottom-up adoption inside large enterprises before it had a mature sales motion.

The Ann Arbor story

Voxel51 was founded in 2016 by Jason Corso, a professor of electrical engineering and computer science at the University of Michigan, and Brian Moore, who had worked with Corso on computer vision research. The founding happened at the moment when the volume of unstructured visual data inside companies was growing faster than anyone's ability to organize it. Corso's research lab at Michigan had been working on exactly that problem for years.

The earliest chapter of the company is part of Ann Arbor ecosystem folklore. Voxel51 got its start inside Ann Arbor SPARK and in an office above Arbor Brewing Company on Washington Street. Moore is CEO. Corso is Chief Scientist and still holds his faculty appointment at Michigan. That faculty link keeps a live research pipeline open between U-M's computer vision and robotics labs and the commercial product. The company has never moved its headquarters out of Ann Arbor.

Funding came in stages. A federal research grant seeded the earliest work. A $2M seed followed. A $12.5M Series A in 2022, co-led by Drive Capital and Shasta Ventures, gave the company the runway to scale the commercial product. The $30M Series B in May 2024, led by Bessemer Venture Partners with participation from Drive Capital, Tru Arrow Partners, Shasta Ventures, and ID Ventures, was the signal event.

The Bessemer round was the thing that reordered how out-of-state VCs looked at Ann Arbor AI. Bessemer is a top-tier coastal firm with deep enterprise infrastructure chops. Leading a Series B in an Ann Arbor AI company validated a thesis that had been half-whispered in the region for years: Ann Arbor can originate and scale a serious AI infrastructure company without relocating its center of gravity to San Francisco.

Culture and hiring

Voxel51 hires machine learning research engineers, software engineers (Python heavily, backend and infrastructure), developer advocates, product managers, designers, and field ML engineers who help customers get their data into shape. The team skews technically deep. Academic credentials are common but not required.

The work profile suits engineers who like open-source-first product development. FiftyOne's community matters to the product, and contributor interaction is real work, not a side channel. Developer relations is a first-class discipline at the company.

Work model as of 2026 is hybrid with an Ann Arbor center of gravity. Remote roles exist for specific functions. The exact in-office cadence varies.

Glassdoor and public employee signal for Voxel51 is limited in volume but trends positive on technical work, mission, and peer quality.

Strategic reading

Voxel51 sits in a well-defined emerging category. Data-centric AI and ML observability tooling has become a meaningful part of enterprise AI infrastructure spend. The competitive frame includes Scale AI (much larger, more labeling-heavy), Labelbox, Snorkel AI, and the data-side offerings inside Databricks, Weights and Biases, and Hugging Face.

Voxel51's distinguishing feature is the combination of open-source core adoption (which large enterprises trust because they can inspect and extend it) and computer vision specialization (which gives the product depth that horizontal tools lack). The strategic question is whether that combination scales into a durable commercial business as the market consolidates.

Risks worth naming. AI infrastructure is a crowded category with well-funded competitors. A few of them have meaningful distribution advantages through existing MLOps platforms. Enterprise deal cycles are long and expensive to run. AV-industry customer concentration is a real exposure; if AV capital continues to tighten, specific customer revenue could be volatile.

Bull case: FiftyOne's open-source footprint is the cheapest customer acquisition channel in the category, the Databricks partnership plugs the company into the largest enterprise data platform outside the hyperscalers, and the Bessemer-led Series B gives the company runway to scale a sales motion without compromising the open-source core.

The broader read on Voxel51 for Ann Arbor watchers: this is the company that answered whether the region could produce and retain a serious AI infrastructure business. The answer so far is yes. Corso has not moved the company to the Bay Area despite the usual post-Series B gravitational pull of coastal investors. Moore has hired locally, built an engineering organization in Ann Arbor, and kept the company's public identity tied to the city. That matters for the next generation of U-M AI spinouts trying to decide whether staying in Michigan is a feasible path to scale. Voxel51 is the proof.

Positive signals to watch across 2026-2027: FiftyOne open-source adoption growth in telemetry that Voxel51 discloses or that can be inferred from GitHub metrics, any further partnership announcements with the major MLOps and data platforms, and the cadence of enterprise customer references published by the company. A Series C in the 2026-2027 window would be the natural next event.

Negative signals to watch: slowing open-source contribution activity, visible churn in senior engineering leadership, or a pattern of losing deal against Scale or Snorkel in head-to-head enterprise RFPs. None of those signals are visible today, but they are the ones that would reorder the strategic read.

If you're considering an offer

Voxel51 is the best bet in Ann Arbor for working on AI infrastructure at a company that is already well past product-market-fit but still small enough that your work matters.

What you get: real technical scope, a serious engineering peer group, and the chance to ship open-source code that runs inside Fortune 500 ML teams. Ann Arbor cost of living against a growing AI infra company's compensation band is a strong purchasing-power proposition. Equity is private-company stock at growth stage, which means real upside if the company keeps compounding.

What you give up: the coastal AI infra alumni network. You give up some of the specific career optionality that flows through Bay Area AI infra companies. Compensation is competitive but will not match late-stage hyperscaler packages.

Compared to other Ann Arbor AI offers, Voxel51 is the most commercially mature of the venture-backed group. Clinc is more vertical-specific and has had a more complicated growth story. MemryX is an edge-AI chip company on a hardware timeline. Genomenon is more specialized in genomics. For a generalist AI infrastructure or computer vision engineer, Voxel51 is the cleanest fit in Ann Arbor.

Other Ann Arbor AI and computer vision companies

  • Clinc - conversational AI for banking, U-M EECS spinout
  • MemryX - edge AI accelerator chips, U-M spinout
  • Genomenon - AI-powered genomic interpretation
  • Invisible AI - computer vision for manufacturing (Ann Arbor office)
  • BlueConduit - AI for water infrastructure and lead service line identification

See the full Ann Arbor AI companies hub.

FAQ

Who founded Voxel51? Jason Corso, a University of Michigan electrical engineering and computer science professor, and Brian Moore co-founded Voxel51 in 2016. Moore is CEO and Corso is Chief Scientist.

How much has Voxel51 raised? Voxel51 has raised approximately $45.5M across a federal grant, a $2M seed, a $12.5M Series A in 2022 co-led by Drive Capital and Shasta Ventures, and a $30M Series B in May 2024 led by Bessemer Venture Partners.

What is FiftyOne? FiftyOne is Voxel51's open-source Python toolkit for building, exploring, and refining computer vision datasets and evaluating the models trained on them. An enterprise tier called FiftyOne Teams adds collaboration, access controls, and integrations.

Where is Voxel51 headquartered? Voxel51 is headquartered in Ann Arbor, Michigan. The company got its start inside Ann Arbor SPARK and in an office above Arbor Brewing Company on Washington Street.

Who does Voxel51 partner with? Voxel51 has publicly announced partnerships with Databricks and with V7 Darwin, the computer vision annotation platform. FiftyOne also integrates with standard MLOps tools, annotation platforms, and cloud object stores.

Who are Voxel51's competitors? Voxel51 competes with Scale AI, Labelbox, Snorkel AI, and the data-side offerings inside Databricks, Weights and Biases, and Hugging Face. Its differentiator is the combination of open-source core adoption and computer vision specialization.

Is Voxel51 hiring? Yes. Open roles are listed at voxel51.com/careers across ML research, engineering, developer relations, product, and field ML roles.

Is Voxel51 profitable? The company does not publish financials. Growth-stage AI infrastructure companies are typically prioritizing growth over profitability at Voxel51's stage.