Tamarind Bio Secures $13.6M Series A to Make AI More Accessible for Biology

The year was 2024. Deniz Kavi, a software engineer at Stanford School of Medicine, recalled the “untenable” research hand-off between wet lab biologists and computational scientists. The rise of models, such as Nobel Prize-winning AlphaFold, which solved the protein structure prediction problem, proposed a powerful inflection point for AI-driven therapeutic development. Yet, much AI tooling was still built for the technical user, requiring command-line access and cloud credentials that bottlenecked implementation in the life sciences lab. 

“Most chemists and biologists are not programmers and can’t write code to use these tools,” explained Kavi in an interview with GEN Edge.  

At Stanford, Kavi found himself in a constant loop with biology lab mates: emails asking to execute AlphaFold runs, results sent back, repeat. The stream of requests exposed a clear need for a centralized platform of model coordination and inference for life science teams. 

Kavi then partnered with Stanford co-founder, Sherry Liu, who carried a cloud computing background from AWS, to build a new AI product for biology that manages overhead infrastructure, including hosting, compute provisioning, runtime orchestration, security, and data residency. The team launched the platform online and saw immediate traction, gaining 600 users in the first month.  

The duo formed a new company, Tamarind Bio, which went forward to raise a seed round from Y Combinator, with Kavi stepping in as CEO.  

Two years later, the team has now announced a $13.6 million Series A led by Dimension Capital. Tamarind’s platform has since expanded to a library of over 200 models covering a wide range of modalities, including antibodies, peptides, small molecules, enzymes, and radiopharmaceuticals. Functional applications span protein design, binding affinity prediction, small molecule generation, molecular dynamics, and more.  

To-date, Tamarind has grown a massive user base of approximately 100 total biotech companies, which includes eight of the top 20 pharma players, such as Bayer and Boehringer Ingelheim. Subscribing partners are given access to the full model library. 

“The best way to focus energy as a computational team at a biotech company is to work on novel science or AI tooling,” Kavi told GEN Edge. “Tamarind handles the cumbersome, non-differentiated parts of that work.” 

Models in bulk 

Tamarind’s launch follows on the heels of a series of 2026 pharma deals that have signaled an AI infrastructure moment.

Nan Li, founder and managing partner at Dimension, said the firm initiated Tamarind’s Series A investment after watching the strong organic adoption within the traditionally conservative pharmaceutical sector, where the stakes are high in both preclinical research and clinical development.  

Additionally, Kavi and Liu proved to be a strong powered founding team that was “better than a 30-person engineering team build.” The duo built and sold the product from scratch without large venture dollars going into sales and marketing. 

“Our team thought that Tamarind had a really compelling vision,” Li told GEN Edge. “As machine learning continues to mature and federate, there would be many models that need to be coordinated and hosted.” 

While the first phase of the AI revolution focused on applications and a hunt for best-in-class models, Li asserts that the field now faces a new problem. Too many models have come out with utility, which now requires consolidated workflows. As an example in a protein preclinical drug discovery campaign, 5-6 models may be used in tandem to generate binders and predict therapeutic properties, such as thermostability, aggregation, and more.

“Our excitement for Tamarind represents this critical next step in how the industry is using machine learning, by adopting models in bulk, not in piecemeal,” Li continued. 

Anecdotes from Tamarind’s partners illustrate the company’s platform capabilities. One anonymized biotech’s data science team was overburdened with genomics, proteomics and computational binder design requests from the bench-focused lead engineering team. The team then implemented Tamarind’s platform access the models for property prediction and molecular design directly.  

The scientists also used the platform to create end-to-end pipelines for discrete, complex tasks in their day-to-day work, including affinity maturation before a high throughput screening assay or filtering leads based on drug-like properties to reduce experimental resources. 

Reflecting on the cultural shift toward AI-powered biology, Kavi is excited to see an increasing number of pure wet lab scientists applying AI models to real problems. “That trend will accelerate,” he says, “our goal is to make these tools ubiquitous for all.”