Dyno Therapeutics Launches Dyno Psi-Phi, an Agentic AI Suite for Protein Binder Design, at NVIDIA GTC 2026

Audio By Carbonatix

WATERTOWN, Mass.--(BUSINESS WIRE)--Mar 18, 2026--

Dyno Therapeutics, Inc., a genetic technologies company applying AI to empower patients everywhere with genetic agency, today announced the launch of Dyno Psi-Phi (pronounced “sci-fi”), a suite of AI-powered protein design tools to help therapeutic developers create sequence-based medicines with the potential to transform patient lives. The announcement took place at NVIDIA GTC 2026 and arose from a collaboration between Dyno and NVIDIA to accelerate therapeutic discovery and design with high-performance computing.

AI technologies that can precisely design the amino acid sequences of protein molecules to bind therapeutic targets have the potential to dramatically accelerate the development of new medicines.

“Dyno’s Psi-Phi design suite is a step toward solving a major challenge in modern AI: connecting in silico benchmarks to real-world outcomes,” said Sam Sinai, Ph.D., Head of Machine Learning and Cofounder at Dyno Therapeutics. “Much of today’s progress is measured against a narrow set of computational filters. While effective, they can limit exploration of the broader functional space of proteins. With Psi-Phi, we democratize the filters that work, while introducing models that generate greater diversity and pair naturally with high-throughput experiments, so designs succeed not just at binding, but across downstream requirements.”

Dyno Psi-Phi is a suite of integrated tools:

Dyno Phi is a collection of filters that strongly predict real-world outcomes. Dyno iteratively calibrates and improves these filters with real-world experimental data to enable users to accurately assess the likelihood that resulting designs will validate in experimental settings. Furthermore, users can easily combine and modify the application of these filters to fit their design needs, balancing confidence in design success with diversity of designs. Dyno Phi can be used with any generative model to increase the confidence of translating designs to a wet-lab setting.

Dyno Psi-1 is the first open-weight model from the Dyno Psi protein design family. It is influenced by the NVIDIA La-Proteina family of models and trained on the NVIDIA DGX Cloud using NVIDIA Hopper GPUs. With a flow-matching backbone generative model optimized for scaling to complex multimeric interfaces, Psi-1 prioritizes structural diversity and controllability over pure in silico scoring objectives, producing more novel binders and demonstrating excellent experimental performance across common lab assays. The next-generation development of these models has already integrated the NVIDIA Transformer Engine to accelerate training and inference. To facilitate broader development of models within the community, Dyno is also openly releasing a curated dataset of synthetic domain interfaces (SDI) on which Dyno Psi-1 has been trained to improve the diversity of designed binders.

Dyno Psi-Phi APIs are a collection of community-tier REST APIs that provide programmatic access to Dyno Psi-1 for binder generation, and Dyno Phi for sequence filtering and experimental grounding. Developers can generate and evaluate protein designs through a unified interface without provisioning GPUs or building custom calibration pipelines, enabling structure-conditioned design and probabilistic estimation of real-world binding performance, and incorporating NVIDIA BioNeMo NIMs such as OpenFold3. These APIs are accessible at design.dynotx.com.

Dyno Psi-Phi Claude Code Skills are a one-click extension of the Psi-Phi platform that provide authenticated access to the same GPU-backed APIs directly within Anthropic’s Claude Code. They invoke inference and experimentally grounded filtering in real time, allowing scientists to iterate on protein design conversationally while leveraging Dyno’s computational infrastructure.

“Our collaboration with NVIDIA opens a new world of extraordinary AI leverage for Dyno’s partners, making frontier AI accessible to an even broader community of therapeutic developers,” said Eric Kelsic, Ph.D., Chief Executive Officer and Cofounder of Dyno Therapeutics. “Many genetic diseases can be treated by modulating protein binding, but historically it has been challenging to control and scale generative AI models for protein binder design. Our close technical partnership with NVIDIA, along with NVIDIA’s advanced hardware and software infrastructure, has enabled us to make rapid progress training Dyno Psi-1, and to create effective filtering methods with Dyno Phi agents. Now therapeutic developers can more confidently apply AI to create exceptionally effective new medicines for the patients who need them most.”

“Bridging the persistent gap between computational workflows and successful real-world experimental outcomes is critical for modern molecular design, and NVIDIA DGX Cloud equipped with H200 GPUs provides developers with the advanced infrastructure necessary to train complex generative models at scale,” said Anthony Costa, Director of Digital Biology at NVIDIA.

Dyno Psi-1 is available at https://huggingface.co/dynotx/dynopsi.

Dyno Phi is available at design.dynotx.com and through Claude Code .

Dyno SDI dataset is available at https://huggingface.co/datasets/dynotx/synthetic_dimers.

About Dyno Therapeutics

Dyno Therapeutics is on a mission to build high-performance genetic technologies that transform patients’ lives. Dyno applies AI to create better technologies for gene delivery and sequence design to increase “Genetic Agency” – a person’s ability to take action at the genetic level to live a healthier life – through safe, effective and widely accessible genetic medicines. Dyno partners across industries to ensure these life-transforming technologies can help as many patients as possible, including through strategic collaborations with leading gene therapy developers Astellas and Roche. Visit www.dynotx.com.

View source version on businesswire.com:https://www.businesswire.com/news/home/20260318369291/en/

CONTACT: Media Contact:

Thermal for Dyno Therapeutics

[email protected]

KEYWORD: UNITED STATES NORTH AMERICA MASSACHUSETTS

INDUSTRY KEYWORD: DATA MANAGEMENT BIOTECHNOLOGY TECHNOLOGY HEALTH GENERAL HEALTH PHARMACEUTICAL RESEARCH SOFTWARE ARTIFICIAL INTELLIGENCE GENETICS SCIENCE

SOURCE: Dyno Therapeutics, Inc.

Copyright Business Wire 2026.

PUB: 03/18/2026 12:00 PM/DISC: 03/18/2026 12:00 PM

http://www.businesswire.com/news/home/20260318369291/en