Novarna Wants AI to Learn the Language of RNA—and Then Prove It at the Bench
The fashionable story in AI drug discovery is software finding a molecule. Novarna is pursuing a harder loop: generate RNA designs, predict their structure and interactions, build them in a wet lab, measure what happened, and use the experimental data to make the next designs better.
CEO and founder
Head of R&D and co-founder
CTO and co-founder
Crystal Brown learned biotech the expensive way
Crystal Brown’s path into biotechnology does not fit the standard founder biography. She built an operations career across large organizations, including automotive manufacturing, before moving into biotech. Her first attempt at building a biotechnology company did not end with a triumphant exit. As Brown later told TechCrunch, the company burned through capital and the personal consequences were severe. Failure became part of the education.
That history matters because Novarna is a company about closing loops between theory and execution. Brown learned the same lesson at the company level that the platform is trying to encode scientifically: an elegant idea is not enough. The physical system has to work. Capital has to last. Experiments have to resolve uncertainty. Operations are not beneath the science; they are how science becomes a company.
CircNova came first
The company began in May 2023 as CircNova. Contemporary reporting identified Brown, scientist Joe DeAngelo and William Grenawitzke as the original founding team. The early thesis centered on circular RNA, a class of RNA molecules whose closed-loop structure can offer attractive stability characteristics compared with many linear RNA designs.
The history is worth preserving because the current Novarna team page tells a later-stage organizational story. Today it identifies Brown as CEO and founder, Alex Wesselhoeft as Head of R&D and co-founder, Jean-Philippe Bürckert as CTO and co-founder, and DeAngelo as Executive Scientific Advisor. Those descriptions can coexist with the original 2023 record: teams evolve, scientific leadership changes, and companies often redefine roles as the technical platform broadens.
Why the company became Novarna
In August 2026 CircNova changed its name to Novarna. The company says the change reflects an expansion beyond circular RNA into the major RNA modalities. That is more than cosmetic positioning. A platform company becomes more valuable when its core capabilities—sequence generation, structure prediction, target interaction modeling and experimental learning—can transfer across classes of RNA rather than remaining trapped inside one molecule type.
The strategic bet is that the reusable asset is not one drug candidate. It is the machinery for designing many RNA candidates and learning from what the lab says about them.
Inside NovaEngine
Novarna describes NovaEngine as a purpose-built AI stack for RNA. Its components span novel sequence generation, two- and three-dimensional structure prediction and modeling interactions between RNA and biological targets. The technical ambition is important because RNA is not simply text with four letters. Biological behavior depends on structure, folding, stability, localization, chemistry and interactions inside noisy living systems.
A model that proposes a sequence without understanding physical constraints can generate impressive-looking nonsense. Conversely, a system that only predicts structures but cannot search design space does not close the discovery loop. Novarna is trying to connect those steps so a target can lead to candidate designs that are more informed before expensive experimental work begins.
The wet lab is the strategic center
The most important part of the architecture may be NovaLab, Novarna’s wet-lab operation. AI-biotech companies live or die by the relationship between computational prediction and biological evidence. A closed loop lets the company design molecules computationally, synthesize or test them, capture the results and feed those results back into future modeling.
That creates the possibility of a proprietary data flywheel. Public biological datasets are valuable, but they are available to competitors too. Carefully designed experiments can generate data that is specific to the company’s models and failure modes. Over time, the unique asset may be less the first generation of algorithms than the accumulated record of what those algorithms got right and wrong in the lab.
An encouraging experiment is not a medicine
Novarna has published a case study from work with a University of Michigan laboratory involving AI-designed RNA candidates tested in ovarian-cancer cell models. The company reports that two designs produced substantially higher on-target efficiency in those experiments than a commercial linear siRNA comparator. That is intriguing evidence that the design system can produce biologically active candidates.
It is equally important to say what the result is not. A cell-model experiment is not a clinical trial, not evidence of safety in people and not proof that a therapy will survive the long path through delivery, toxicology, manufacturing and regulation. The disciplined way to read the result is as early validation of a design platform—not as a cancer-treatment claim.
The scientists behind the expansion
Alex Wesselhoeft brings unusually relevant RNA depth. His work has included circular RNA engineering, RNA therapeutics research and technology development across MIT, Harvard-affiliated research and industry. He previously helped build Orna Therapeutics around circular RNA technology; Lilly agreed in 2026 to acquire Orna in a transaction valued at up to $2.4 billion. That does not transfer success automatically to Novarna, but it signals the level of domain experience entering the company.
Jean-Philippe Bürckert’s role as CTO places the software and computational architecture alongside that experimental expertise. In AI-biotech, the organizational problem mirrors the scientific one: machine learning cannot be a detached software team handing predictions to biologists. The useful system is built where computational scientists and experimental scientists argue with the same data.
A business model built around discovery
Novarna is positioning the platform for multiple engagement models rather than betting the company entirely on one internal therapeutic pipeline. Its commercial language includes rapid-discovery projects, optimized-lead programs and longer discovery partnerships. That can reduce capital intensity relative to owning every downstream clinical program while still allowing the company to capture value from its platform.
The trade-off is strategic focus. Service-heavy models can generate revenue and data but can also turn a platform company into a custom research shop. The long-term question is whether Novarna can use partnerships to make the platform stronger while retaining enough proprietary upside to justify the cost and risk of deep technology.
What has to go right next
The company raised a $3.3 million seed round in 2025 as CircNova. For a biotech platform with a wet lab, that is meaningful but not enormous capital. The next stage will demand evidence across more targets, more RNA modalities and more independent experimental settings. Model performance must transfer. Laboratory results must reproduce. The company must show that its design loop produces better economics or better molecules—not merely faster slide decks.
If Novarna can do that, the company has a compelling place in the emerging AI-biotech stack. It is not trying to make a general language model speak biology. It is trying to build a specialized learning machine around RNA and force that machine to confront reality in the lab. That is a difficult thesis. It is also exactly why the company is worth watching.
Sources
- Novarna — technology platform and current team
- Novarna — CircNova is now Novarna
- TechCrunch — CircNova founding history, seed round and NovaEngine
- Mass General Brigham — Alex Wesselhoeft RNA-therapeutics profile
- Novarna — published ovarian-cancer cell-model case study
- Black Enterprise — Crystal Brown’s biotechnology path