Cellframe raised $76 million on Monday in a Series B led by Lux Capital, with participation from Founders Fund and the venture arm of Thermo Fisher. The Austin company sells software that orchestrates robotic biology laboratories: scheduling instrument time across competing experiments, handling the constant small failures that automated wet lab work produces, and recovering a run without losing the samples. It is not a glamorous product and it addresses the reason a large share of expensive laboratory automation sits underused. Twenty-two customers run it, including three of the ten largest pharmaceutical companies and several academic core facilities.
Why automated labs underperform
A modern biology lab may contain liquid handlers, plate readers, incubators, centrifuges, and a robotic arm to move plates between them, from four or five different manufacturers with incompatible control software. Running an experiment requires choreographing all of it with timing constraints, because cells do not wait. If any instrument reports an error partway through, and they do constantly, the run either aborts and wastes reagents costing thousands of dollars or a human intervenes in real time.
The result across the industry is that expensive automation gets used for the simplest and most repetitive protocols, where failure is cheap, and complex experiments stay manual. A laboratory operations director at a large pharmaceutical company told us her utilization on a $2.4 million automated system runs around 30 percent, and that the constraint is not instrument time but the engineering effort to make each new protocol reliable.
What Cellframe built
A scheduler that models each instrument's capabilities and timing, plus a protocol description language where a scientist specifies constraints rather than a sequence. The system computes a schedule, executes it, and, critically, replans when something fails. If a plate reader throws an error, the scheduler can hold the plate in an incubator within its tolerance window, retry, and route to a second reader if one exists, all without discarding the experiment.
The replanning is the hard part and it is where the company's engineering has gone. Chief executive Priya Raghavan, previously at Ginkgo Bioworks, described the core insight as treating biological tolerance windows as scheduling constraints rather than as fixed times. A step that says incubate for 30 minutes usually means anywhere between 27 and 40, and knowing that slack is what allows recovery. Encoding it requires domain knowledge, which the company gets from a scientific staff that outnumbers its sales team.
The story is rarely the launch. It is what breaks, what ships, and who owns the mess at 2 a.m.
The measurable result
Customers report run completion rates rising from a typical 60 to 70 percent to above 92 percent, and instrument utilization roughly doubling. One academic core facility at a Midwestern university shared specific numbers: 41 percent utilization before, 79 percent after, with the same staff running 2.4 times as many experiments. For a facility with a capital budget that will not grow, that is equivalent to a second lab.
The pharmaceutical customers care less about utilization and more about reproducibility. A run that recovers from a failure in a documented, deterministic way produces data an auditor accepts. One customer said the compliance argument closed the purchase, not the throughput one, because their quality organization had been blocking automation for exactly this reason.
The market and the competition
Instrument manufacturers each ship their own orchestration software that works well with their own hardware and poorly with anyone else's, which is a deliberate commercial choice. Thermo Fisher investing in Cellframe while selling competing software is a hedge that the company acknowledged in a brief statement. Emerald Cloud Lab built a fully automated lab as a service and has struggled commercially. Several academic open source projects exist and none has the reliability engineering that production use demands.
The competitive risk is that the Standardization in Lab Automation consortium, which has been working on interoperability specifications for years, eventually produces something that commoditizes the integration layer. Raghavan's view is that standards would help her rather than hurt, because integration is the cost of doing business and the scheduling intelligence is the product. That is probably right and it depends on the standards actually arriving, which lab automation standards have a poor record of doing.
What the funding buys
Mostly integration engineering. Each new instrument model requires work, and the company supports about 140 today against a market with several thousand. Raghavan said the roadmap prioritizes by customer request and that the pace is roughly six new integrations a month with a team of eleven. Doubling that team is the largest line in the plan.
The second priority is a protocol library, a set of validated automated workflows customers can adopt rather than write. That moves the company from selling infrastructure to selling content, which changes the sales conversation and creates a compounding asset. It also invites the objection that scientific protocols are institution-specific, which Raghavan disputes for the common assays and concedes for anything novel.
The risk nobody at the company raised and two customers did is that better scheduling exposes the next constraint rather than removing it. A facility running at 79 percent instrument utilization discovers that its sample preparation, its reagent ordering, or its data analysis pipeline is now the bottleneck, and none of those are things Cellframe sells. That is the ordinary experience of removing a constraint from a system and it means the second year of a deployment produces less visible improvement than the first. Customers who understood that going in have been the happier ones.
Skarvonix will keep following this beat with reporting grounded in how systems behave outside the launch keynote.
- Funding
