Vijay Pande: Focused AI Investments Require Open Biological Data

Vijay Pande’s VZVC will make about five investments a year, while Pande expects biology to develop shared datasets for AI

On August 29, Pande discussed VZVC in a TechCrunch interview. He founded the firm with Zak Verner after spending more than a decade at the venture capital firm a16z. VZVC plans to make about five investments a year and assign some of its daily work to AI agents instead of hiring associates. In the same conversation, Pande explained why biological models need data that cannot be obtained from the open internet.

Before founding VZVC, Pande spent more than a decade building a16z’s biomedical practice, which grew to manage nearly $4 billion. He and Verner will handle investments at the new firm. The AI agents they developed, which are programs that can perform some daily tasks, have removed the fund’s need to hire associates.

The partners plan to make about five investments a year instead of thirty and work with founders for five to ten years or longer.

Before becoming an investor, Pande created Folding@home, a distributed computing project for disease research. In the same conversation, he identified another constraint on the use of AI in medicine. Language models are trained on vast collections of publicly available text. Biological data cannot be collected from the internet in the same way. Nearly every company builds its own private dataset, so no shared body of training data emerges.

The Israeli consortium IBFI is already trying to address a similar problem. Its purpose is to standardize data from hospitals, laboratories, and companies so that models can compare these sources with one another.

Pande expects a shift toward biological information atlases and open foundation models, meaning large AI models that can subsequently be applied to different tasks. He compares their potential effect with that of open language models, which compete successfully with proprietary models.