How biotech ML teams scale reproducible, auditable workflows as data and compute demands grow

Description

AI-native biotech teams have made machine learning central to how they do discovery and diagnostics, and the work behind it keeps getting heavier: terabyte imaging, multi-step genomics, molecular simulation, GPU-heavy training, and experiments that can take days or months to finish. More of it runs on its own now, with longer chains of steps and less hands-on supervision. Most teams start with their own scripts or a basic scheduler, and that works for a while. Then something shifts. A new data type, a bigger team after a raise, or the first real conversation with regulators. Suddenly you can't re-run an experiment from six months ago without guessing at what changed. Compute bills go up and no one can explain where the money went. The audit trail that used to be fine isn't anymore, and for teams working with patient data, privacy and residency rules like HIPAA and GDPR start shaping how the infrastructure gets built. The team has to decide whether to keep extending what they have, standardize on a platform, or rethink the architecture. This roundtable puts ML and engineering leaders from AI-native biotech and life-sciences companies in the same room to talk through that shift, from reproducibility and compute cost to compliance and keeping research teams focused on the science as the work scales.

Speakers

Rody Arantes

Director of Digital Technology, Montai Therapeutics

Rick Sullivan

VP of Delivery Sciences, Shape Therapeutics

Niels Bantilan

Chief Machine Learning Engineer, Union.ai

Tanu Malik

Associate Professor of Electrical Engineering & Computer Science, University of Missouri

Roland Scharrer

Partner and Managing Director, Kearney

Guided Questions

Tanu Malik

Your lab builds the tools that let one scientist reproduce another's computational work, and you've measured how often that actually fails in practice, in one recent study, roughly a third of AI-generated projects wouldn't even run, usually from hidden dependencies. From everything you've measured, what's the real gap between a result that looks reproducible and one that actually is?

Rick Sullivan

At Shape you build the ML and the wet-lab work as one system, so a model's prediction and what actually comes back from the bench are constantly shaping each other. Having built it that way from the start, what does that let you do that teams who treat the two as separate stages never quite get to?

Roland Scharrer

You've led data and AI strategy across a large, complex organization, and your research looks at why some turn AI into real advantage while others stall. Across all of it, what separates the teams that get lasting value from AI from the ones that get stuck, and how much of that comes down to the foundations they build rather than the models themselves?

Rody Arantes

At Montai you built a framework where any scientist can clone a repo, work with an LLM, and stand up a real, deployable app, security and guidelines already handled. You've got scientists shipping full apps other scientists now use. Having watched that happen, what did they do with it that you didn't expect, and what did you have to build in so their apps stay reliable without your team checking every one?

Niels Bantilan

You've spent years building the orchestration layer that pipelines run on, and lately it's colliding with AI that doesn't behave like traditional pipelines, non-deterministic, retrying, kicking off its own steps. What does that break about the assumptions orchestration was built on, and what has to change underneath to run it reliably?