​Description

​This session brings together researchers and engineers working across three layers of the problem: representation learning that puts molecules and proteins in one searchable space, genome language models that predict, design, and interpret sequence, and agentic tooling that turns scattered evidence into experiments, plate layouts, and documentation.

​This event is designed for practitioners in computational biology, drug discovery, and applied ML who want to understand what is already viable in silico and where wet-lab validation is still the bottleneck.

​​Agenda

​We start at the model layer: how sequence and structure become embeddings a system can search. From there we move up to genome-scale modeling and what post-training buys you on real clinical benchmarks. The final talk closes the loop, taking retrieved and ranked evidence through to bench work, scripting, and documentation.

​18:00 Doors open

​18:30 - 19:00 A new frontier in generative genomics with Omnii (Stefano Massaroli, Co-founder @ Radical Numerics)

​19:00 - 19:30 Agentic AI for the Wet Lab (Daniel Leuck, AI Engineer @ Ikayzo, and Calvin Duong, Researcher @ University of Tokyo)

​19:30 - 20:00 Molecule and Protein Embeddings with DrugCLIP (Ewa Szyszka, DevRel Engineer @ Qdrant)

​20:00 - 21:00 Networking

​21:00 Doors close

​Talks

​Talk 1 - A new frontier in generative genomics

​Abstract: Omnii is a research preview of a genome language model built to predict, design, and interpret. It is pretrained with native fusion of genomic annotation tracks alongside DNA sequence and post-trained so it can be used without training task-specific probes on frozen embeddings. This talk covers the architecture behind it, including block convolutions and sparse attention over a 2 million base pair context window, the benchmark picture across clinical variant interpretation and complex traits, and what it looks like in practice: recovering experimentally validated functional variants in Alzheimer's microglia loci from sequence alone, and chain-of-thought design of RNA aptamers.

​Bio: co-founder and president of Radical Numerics Inc., a Research Scientist on RIKEN's Deep Learning Theory team, formerly a founding scientist at Liquid AI where he helped launch Liquid AI Japan, a postdoc under Yoshua Bengio at Mila, co-inventor of hybrid convolution language models and neural differential equations, and a University of Tokyo PhD — presented "Design Principles for Training at Extreme Context Lengths."

​Talk 2 - Agentic AI for the Wet Lab

​Speaker: Daniel Leuck (AI Engineer, Ikayzo) and Calvin Duong (Researcher, University of Tokyo)

​Abstract: In silico approaches to faster, smarter, and more comprehensive discovery, automation, scripting, and documentation

​In biology, important discoveries are rarely isolated facts. They often emerge from connections between genes, proteins, pathways, experiments, and prior observations. Interaction maps help researchers organize these connections, but the evidence needed to build them is often scattered across papers, databases, and underused sources such as non-English theses that conventional search can miss.

​The future wet lab will need more than better search. As software and hardware become more connected, labs will need systems that turn scattered evidence into action: candidate prioritization, assay suggestions, validation plans, plate layouts, labels, user scripting, and rich documentation that captures work as it happens.

​Using influenza research as a case study, we will introduce Reveria (formerly LabNexus), a platform (beta) for AI-assisted wet-lab work built around modular CodeLets. One example is Discovery, a CodeLet that helps retrieve, translate, parse, and connect fragmented scientific evidence into traceable interaction maps and ranked candidate cards for downstream experiments. The goal is practical: help researchers understand what is already known, decide what is worth testing next, and turn scattered knowledge into better wet-lab experiments.

​Bios:

• ​Calvin Duong is a postdoctoral researcher at the University of Tokyo’s Institute of Medical Science studying influenza immunity and host responses to infection. His recent co-first-author Cell paper used CRISPR-engineered mice to uncover host factors involved in influenza infection (link). The work helped inspire Reveria Discovery, an AI-assisted platform for turning scientific literature into structured evidence and experimental hypotheses, alongside his broader interest in using AI to accelerate antibody and drug discovery.

• ​Daniel Leuck is CEO and co-founder of Ikayzo, a boutique software development and digital design firm with offices in Tokyo and Honolulu. He currently focuses on Reveria AI and RevNexus, a suite of applications for analytics, automation, and scientific discovery. Daniel has 20+ years of technology leadership experience, including Global Head of Development at LastMinute.com and SVP of R&D at ValueCommerce. He has extensive finance and fintech experience with Bank of America, Wells Fargo, PIMCO, Nomura, and hedge funds.

​Talk 3 - Molecule and Protein Embeddings with DrugCLIP

​Speaker: Ewa Szyszka (DevRel Engineer, Qdrant)

​Abstract: DrugCLIP places molecules and proteins in a shared embedding space, which turns "which compound binds this target?" into a vector search problem. This talk walks through how those embeddings are produced, using ESM-2 for protein sequence, how drug and protein candidates are indexed and screened at scale, and how to inspect the resulting space. Screening happens virtually; the wet lab is reserved for validating what comes out at the top.

​Bio: TBD

​​Organizers

​Ilya Kulyatin is an entrepreneur with work and academic experience in the US, Netherlands, Singapore, UK, and Japan. He holds a BA in Economics, an MA in Finance, and an MSc in Machine Learning. He's a 3x founder, now helping Japan grow the local AI ecosystem through a not-for-profit community, Tokyo AI (TAI), while building an AI-native system integrator and solutions provider, Foundry Labs株式会社.

​Ewa Szyszka is ... to fill

​​Supporters

​Foundry Labs K.K. is a Tokyo-based AI systems integrator and solutions provider, delivering end-to-end support for enterprises: from strategy design through implementation, deployment, and operations. They tailor AI to each client's operational, regulatory, and security requirements, with hands-on experience across finance, government, and industry, and a track record of shipping production systems in secure and regulated environments.

​Qdrant is ... to fill

​About TAI

​Tokyo AI (TAI) is the largest international AI community in Japan, with 5,000+ members mainly based in Tokyo: engineers, researchers, investors, product managers, and corporate innovation leaders. Through 80+ events a year and 300+ speakers spanning startups, enterprises, and academia, TAI connects the people building AI in Japan with the global ecosystem, working to transform Tokyo into a global AI hub.

​​​Privacy Policy

​We will process your email address for the purposes of event-related communications and ongoing newsletter communications. You may unsubscribe from the newsletter at any time. Further details on how we process personal data are available in our Privacy Policy.