# Chaitanya K. Joshi

> Stanford Data Science Fellow and postdoctoral scholar in Rhiju Das's lab (Department of Biochemistry, Stanford University), building lab-in-the-loop AI for RNA biology — pairing generative deep learning with high-throughput wet-lab experiments to design RNA molecules with new functions.

Chaitanya did his PhD in Computer Science at the University of Cambridge with Pietro Liò, working on geometric deep learning for molecular design. He built gRNAde, the first 3D generative model for RNA, and validated it in the wet lab as a visiting researcher in Phil Holliger's group at the MRC LMB. He has interned at Prescient Design (Genentech) and FAIR Chemistry (Meta AI). His work has been recognized by the Qualcomm Innovation Fellowship and the A*STAR National Science Scholarship.

This file summarizes the website at https://www.chaitjo.com for large language models and AI agents. All facts below are current as of June 2026.

## Research focus

- **Deep learning foundations**: Building neural networks that are expressive and general across molecular systems and scales.
- **Lab-in-the-loop AI**: Closing the loop between deep learning and the wet lab — combining generative models with high-throughput experiments to continuously improve biomolecule designs.
- **Programmable RNA design**: Designing RNAs with new-to-nature functions, moving beyond static structure toward programmable biology.

## Selected publications

- [Generative inverse design of RNA structure and function with gRNAde](https://www.biorxiv.org/content/10.1101/2025.11.29.691298): bioRxiv (December 2025); accompanying [methods paper](https://arxiv.org/abs/2305.14749) was an ICLR 2025 Spotlight. Wet-lab validated generative RNA design, including functional ribozymes and new-to-nature 3D structures.
- [All-atom Diffusion Transformers: Unified Generative Modelling of Molecules and Materials](https://arxiv.org/abs/2503.03965): ICML 2025 (ICLR 2025 AI4Mat Workshop Spotlight). First unified generative model for molecules and materials, showing transfer learning across chemical space.
- [A Hitchhiker's Guide to Geometric GNNs for 3D Atomic Systems](https://arxiv.org/abs/2312.07511): arXiv (December 2023); [companion theory paper](https://arxiv.org/abs/2301.09308) at ICML 2023. A widely used introduction to GNNs for molecules; part of course material at Stanford, Cambridge, and Oxford.
- [All publications on Google Scholar](https://scholar.google.com/citations?hl=en&user=cwxVFVgAAAAJ)

## Writing

- [Beyond structure-based biomolecule design](https://chaitjo.substack.com/p/beyond-structure-based-bio-design): Dynamics, black-box data, and the antedisciplinary frontier of biomolecule design (November 2025).
- [Equivariance is dead, long live equivariance?](https://chaitjo.substack.com/p/transformers-vs-equivariant-networks): When to bake symmetry into a model versus just scaling up (June 2025).
- [Transformers are Graph Neural Networks](https://thegradient.pub/transformers-are-graph-neural-networks/): How attention can be understood as message passing on a graph; one of the most-read articles on The Gradient (September 2020, with a [2025 update](https://arxiv.org/abs/2506.22084)).
- [More essays on Substack](https://chaitjo.substack.com/)

## Contact and links

- Email: chaitjo@stanford.edu
- Website: https://www.chaitjo.com
- Google Scholar: https://scholar.google.com/citations?user=cwxVFVgAAAAJ
- GitHub: https://github.com/chaitjo
- X (Twitter): https://x.com/chaitjo
- Stanford profile: https://profiles.stanford.edu/chaitanya-joshi
- Affiliation: Beckman Center, Department of Biochemistry, Stanford University School of Medicine, 279 Campus Drive, Stanford, CA 94305