OpenProtein.AI opens AI protein-design tools to working biologists

OpenProtein.AI is trying to remove one of the main barriers where AI meets biology. Powerful protein models already exist, but most researchers still find them too hard to use without machine learning skills, GPU access and a dedicated engineering team.

The startup was founded by MIT graduate Tristan Bepler and former MIT associate professor Tim Lu. Their idea is simple: give biologists a no-code web platform where they can upload their own data, run protein engineering models, predict the structure and function of molecules, and train and fine-tune models for specific tasks. The company already works with pharmaceutical and biotech organizations of various sizes, and offers the platform free to academic scientists. In effect, this is not one narrow tool but a full working toolkit for research, in which AI becomes part of the lab workflow rather than a separate experiment for data scientists.

The project grew out of Bepler's academic work at MIT. While studying in the computational and systems biology program under Professor Bonnie Berger, he worked on a question that remains central to the whole field: how best to understand the relationship between a protein's sequence, its structure and its function. Even before AlphaFold appeared, Bepler was studying how to use evolutionary data to predict protein properties, and eventually arrived at one of the early generative models of this class, essentially a protein language model. The logic was not only to predict a molecule's shape but also to move faster from sequence to an understanding of what exactly that protein can do.

Later, after defending his PhD in 2020, Bepler joined Tim Lu's lab as a postdoc. There the gap between cutting-edge AI tools and the real needs of biologists became especially visible. The models were getting stronger, but using them demanded too much technical preparation: writing code, configuring compute, assembling sequence libraries, fine-tuning, and interpreting results.

OpenProtein.AI is built as a response to exactly this problem. Instead of forcing researchers to become ML engineers, the company moves the complexity into infrastructure and leaves users a clear interface and ready-made workflows.

OpenProtein's key in-house development is PoET, the Protein Evolutionary Transformer. It was trained on groups of proteins so that it can generate related sequences and capture the evolutionary constraints that determine molecular properties. The company claims PoET can generalize these constraints and take in new sequence information without full retraining, which matters especially for labs that constantly produce fresh experimental data. Researchers can use their own data to train models and optimize protein sequences, then run the resulting variants through analysis tools, structure predictors and other in silico checks before moving to wet-lab work. The platform also has an API for those who need programmatic access, but the core workflow remains no-code.

The company continues to expand the platform. In 2025 it introduced PoET-2, a new version of its protein language model that, according to OpenProtein, outperforms substantially larger models while requiring only a fraction of the compute and experimental data. That matters not only for quality but also for research cost: if effective models become lighter, they can be used by more than just the largest pharmaceutical companies.

Big business is already on board, too. Boehringer Ingelheim started using the platform in early 2025 and then expanded its collaboration with OpenProtein to cover protein engineering for cancer therapy and for autoimmune and inflammatory diseases.

The company's next step is to teach its models to handle not only the static properties of proteins but also their dynamic behavior. This covers cases where a protein must take part in several biological mechanisms at once, or change its function after binding to another molecule. If AI can describe and design such scenarios, it will widen the range of therapeutic approaches and make the design of biological systems noticeably more precise.

The main point of this story is that the market is gradually moving from rare, one-off AI experiments in biology toward more accessible infrastructure for everyday research. If OpenProtein.AI manages to balance openness, usability and model quality, it will lower the entry barrier for labs, speed up hypothesis testing and shorten the path from computational protein design to a real candidate for therapy or industrial use. And, perhaps just as important, it would keep the strongest AI tools from being locked inside a handful of large players.