Terray Therapeutics

With @nvidia's Nemotron 3.5 Super VL and previous versions we have embedded the ability to extract chemical synthesis routes from scientific literature. By embedding this capability directly into our EMMI platform, we are enabling our AI agents to dynamically parse and reason over chemical documents on-the-fly, and powering a continuous data flywheel that is further expanding our access to actionable chemical space and accelerating the path to novel medicines for the patients who need them. Thank you to @Nvidia and the team at @NVIDIAHealth! #AI #DrugDiscovery #HumanHealth

Accurate extraction of chemical synthesis routes from scientific literature is critical to accelerating AI-native drug discovery workflows. Today, in collaboration with @nvidia , we are excited to announce state-of-the-art performance in extracting these routes using fine-tuned Nemotron 3.5 Super VL and Nemotron 3 Nano Omni. We are grateful to the team at @NVIDIAHealth for their partnership on this work and honored for our collaboration to be featured in today’s Nemotron 3.5 Super VL launch! By embedding this capability directly into our EMMI platform, which combines proprietary hardware, scale data, chemistry-first AI, and lab-in-the-loop automation, we are enabling our agents to dynamically parse and reason over chemical documents on-the-fly, and powering a continuous data flywheel that is further expanding our access to actionable chemical space and accelerating the path to novel medicines for the patients who need them. Read our blog: Enhanced Parsing of Chemical Reaction Diagrams terraytx.ai/news-and-researc… Read NVIDIA's blog: Nemotron 3.5 Super-VL Powers Specialized Agents with Multimodal Reasoning developer.nvidia.com/blog/nv…

Congratulations to Narbe Mardirossian on being named to @biospace's 2026 40 Under 40 list, recognizing leaders for their innovation, impact, and commitment to improving outcomes for patients.  Over the last six years, Narbe has built the #AI and data backbone for EMMI, Terray’s integrated experimentation and chemistry-first AI platform for novel drug discovery and development.  In just one example, under his leadership, the EMMI platform has generated more than 20 billion experimentally derived binding measurements, accelerating hit identification and powering the entire drug discovery process at Terray. Congratulations Narbe!

Modern structure-based drug discovery heavily leverages an operation known as triangle attention. This O(n³) algorithm is a key component and the dominant cost of TerraBind, the industry’s most efficient binding affinity and structure prediction model, which we developed at Terray. We benchmarked @nvidia 's new BioNeMo Inference Runtime kernels (CuTeDSL) on our production architecture on H100 GPUs: up to 7.1x faster than standard PyTorch attention, and up to 1.8x faster than our current optimized kernel stack (cuEquivariance), with the advantage growing at the larger crop sizes that matter most for protein–ligand complexes. Integration was straightforward. We had kernels running and benchmarked within a day of getting access -- module-level swaps, no pipeline rewrites required. Congrats to the @NVIDIAHealth team on the launch! #DrugDiscovery

Terray to speak at September Investor Conferences

Great to be at the AI + Bio Summit today and share Terray’s vision with @NYSE TV. Check out Eli Berlin’s interview that hit from the exchange floor. Thanks to NYSE, @decodingbio and Arkea Media Group for hosting this incredible event. Terray Therapeutics Co founder Eli Berlin Explains Approach to AI Drug D... youtu.be/t2ddpFhcx28?si=vvco… via @YouTube

Thank you Seoul! Last week Terray was fully immersed in ICML * Terray researchers Matteo Rossi and Ryan Pederson, PhD presented TerraBind, a universal cofolding-based potency model within Terray's EMMI platform that delivers 20% higher accuracy at dramatically lower compute cost unlocked by taking a fundamentally different, potency-first approach to building one of our core drug discovery models. * Terray also co-hosted an amazing evening with our friends at @NVIDIAHealth, @proximabio and @DCVC , catching up with friends and collaborators.

Last week we shared some of our recent work on TerraSynth, a generative synthesis planner that addresses the synthesis bottleneck in small-molecule drug discovery. While generative models are good at proposing molecules with desirable properties, they often propose molecules that are difficult to synthesize in practice, slowing down DMTA cycles. TerraSynth makes synthesis planning a native part of the design loop rather than a problem to solve afterward. Integrating TerraSynth into our EMMI platform has helped us focus on the molecules that are worth making to advance our programs. Thanks to Zack and @mileswhen for their incredible work on this.

We are thrilled to announce that Terray’s drug discovery platform, EMMI (Experimentation Meets Machine Intelligence), has been recognized as a World Changing Idea by Fast Company! EMMI seamlessly integrates AI and experimentation to explore the dark areas of chemical space to find and optimize novel small-molecule drugs for some of medicine’s toughest challenges. EMMI is transforming drug discovery and is changing the world! Read about EMMI terraytx.ai/news-and-researc… #DrugDiscovery #AI #FCWorldChangingIdeas

We welcomed Lisa Decker, Chief Business Officer, to our team today! Lisa brings years of experience in building meaningful partnerships to Terray. Please join us in welcoming her! #AI #DrugDiscovery

Today, we are excited to announce TerraSynth, the latest addition to Terray's EMMI platform. Synthesizing AI-generated molecules is a key bottleneck in modern small-molecule drug discovery. We built TerraSynth to fix that. TerraSynth is a generative synthesis planner that proposes new molecules together with a recipe for making them, keeping the DMTA cycle moving efficiently. *TerraSynth covers 67% more chemical space than the next best synthesis planner, at 1/1000th of the cost. *TerraSynth is fast to inference, and can be run directly inside the design loop, ensuring that every proposed compound has been optimized for synthesizability. *TerraSynth has prospectively reduced synthesis time for molecules synthesized at Terray by up to 75%. TerraSynth is being used in real programs at Terray every day. To read more about how we built this model and the key results, check out the blog terraytx.ai/news-and-researc…

Partnerships are core to Terray - from our collaborations with Bristol Myers Squibb, Gilead Sciences, and Calico Life Sciences to our co-development work with Odyssey Therapeutics. On Wednesday, April 22 at UNLOCK AI Science in San Francisco, our Co-Founder Eli Berlin will join the panel - "Dealmaking and Partnerships: Who's Actually Writing the Checks" - to explore what it really takes to build partnerships around AI-native drug discovery platforms. He'll be joined by Jonah Cool (Anthropic), Sarah H. Nam (AbbVie), Stacy Feld (Johnson & Johnson), and Chris Leiter (Atria Ventures) for what will be a candid conversation about how the partnership landscape for AI-native platforms is evolving.  If you're interested in dealmaking at the intersection of AI and life sciences, we hope to see you on the 22nd! @EliBerlin3 @JCoolScience @chrisleiter_

Benchling's latest Transcribed episode dives deep into how Terray is wiring hardware, data, and AI into a single engine for small-molecule drug discovery. Our CEO and founder, Jacob Berlin, sat down with Benchling’s founder, @sajithw, to discuss the importance of pairing precise, proprietary data at scale with industry-leading AI models – and most importantly, having it all deeply integrated with workflows in the lab. EMMI takes these elements and turns them into agentic, end to end workflows. They cover why models alone aren’t a moat, how EMMI Select tackles “AI abundance”, and why synthetically constrained generative models are the only practical future. If you care about building real medicines at the intersection of experimentation and machine intelligence, please listen and subscribe! Thanks to the @benchling team! Available on all podcast platforms and YouTube! benchling.com/transcribed/ja…

Join our incredible team!

Terray Therapeutics is heading to ACS Spring 2026. Our co-founder and CEO, Jacob Berlin, Ph.D., will speak at the Presidential Event on “Human, Computer and Artificial Intelligence,” where he’ll share how a chemistry-first, AI-native approach can reshape small molecule discovery. If you’ll be at ACS Spring 2026, join the session to hear what we’ve learned so far—and what’s next—for AI in drug discovery. #ACS150 #DrugDiscovery

Last week we announced a big leap forward for universal potency models with EMMI Predict: TerraBind. TerraBind delivers best-in-class potency prediction while running 26x faster and reducing inference cost by 96% relative to standard industry potency models. A key driver of this performance is our collaboration with @nvidia and the integration of the NVIDIA cuEquivariance library. By centering its computational core on triangle attention and multiplication, TerraBind's architecture is uniquely positioned to fully leverage cuEquivariance's GPU-optimized kernels, providing a 3-4x speedup compared to standard implementations. With EMMI Predict: TerraBind at the center of our Generate-Predict-Select workflow, this combination of proprietary data, purpose-built models, and NVIDIA-accelerated hardware is transforming how we discover and optimize small-molecule therapeutics. We're excited to keep pushing the boundaries of predictive chemistry with partners like NVIDIA as we work to bring better medicines to patients faster. Read our blog terraytx.com/insights/emmi-p… Read our arxiv paper arxiv.org/abs/2602.07735