Your Impact at LILA

Lila is redefining the future of biomedicine by combining large-scale automated data generation with scientific superintelligence. We are building the loop where AI, automation, and experimental biology co-evolve.

We are seeking a Senior ML Scientist to connect that work to human medicine, building the systems that assess whether a clinical program's biology holds up.

Three questions define the work. Is the mechanism well supported? What has actually been established about how this intervention is meant to work, and what has only been assumed. Does the mechanism operate in patients? Human genetics, expression, cohort and prior-trial evidence all bear on whether the biology that works in a model system is present, and rate-limiting, in the population being treated. Is the trial built for that mechanism? Endpoints that read out the right thing on the right timescale, biomarkers that measure what the mechanism actually does, and enrollment criteria that select patients in whom it is operative — or, very often, none of these.

You will not answer these program by program yourself. You will build the systems that do it, drawing on the mechanistic models and structured biological evidence the rest of the group generates, and grounding them in human data. Your own judgment is the specification and the standard those systems are held to, and you will build the evaluations — including forecasts of real program outcomes scored against what was knowable at the time — that tell you whether they are any good.

What You'll Be Building

• Build systems that assess mechanistic support for a clinical program. Take the structured biological evidence and mechanistic models generated elsewhere in the group and turn them into an assessment of whether an intervention's proposed mechanism is established, assumed, or unexamined.
• Ground mechanism in human data. Analyze human genetic, expression, cohort and trial-derived evidence to determine whether a mechanism is present, active and rate-limiting in the relevant patient population — and how heterogeneous it is across that population.
• Assess trial design against mechanism. Evaluate endpoint choice and timing, biomarker definition and assay, dose and schedule, and eligibility and enrichment criteria for alignment with the proposed mechanism, and encode that assessment so it can be applied at scale rather than case by case.
• Build the evaluations that hold these systems to account, including outcome-verifiable forecasts of real program progression scored using only information available at the prediction date, with the evidence boundary enforced against contamination. Run them yourself and report honestly when a contribution adds nothing.
• Co-design with ML scientists, mechanism scientists and engineers, translate model output into decisions people actually make, and publish — what is and is not predictable from mechanistic and human evidence is a real scientific question we intend to answer in public.
• Communicate findings clearly to technical and cross-functional audiences, including scientists, engineers, product partners, and therapeutic stakeholders.
• Support external scientific visibility through publications, presentations, and engagement with ML/AI for Biology, computational biology and therapeutic discovery communities, as appropriate.

What You'll Need to Succeed

• PhD in a computational discipline — translational bioinformatics, computational biology, biomedical informatics, biostatistics, epidemiology, machine learning, or related — with research centered on human biomedical data.
• Hands-on ML and data analysis for translational medicine. Fluent Python; substantial experience analyzing human genetic, multi-omic, cohort, trial or real-world data in reproducible pipelines. You will run your own analyses and evaluations and interpret them yourself.
• Mechanistic reasoning about therapeutic interventions. Able to state how an intervention is meant to work, what evidence would establish each step, and where human evidence supports or undercuts it.
• Clinical development fluency. Working knowledge of trial design, endpoints, biomarker strategy, eligibility and enrichment — enough to read a protocol and judge whether it tests the mechanism it claims to.
• Evidence judgment and a systems instinct. Able to reason about what was knowable when and resist hindsight — and interested in making that judgment reproducible by something other than you, through structure and evaluation rather than case-by-case expertise.

Bonus Points For

• Experience with human genetics for target identification and validation — common and rare variant evidence, QTL and expression data, or genetically supported target work.
• Experience with biomarker development, patient stratification, companion diagnostics, or enrichment strategy.
• Experience with clinical trial datasets, real-world data, or observational cohort analysis, and their known limitations.
• Experience evaluating language models or agents on scientific judgment tasks, including contamination and memorization controls.
• Familiarity with survival analysis, competing risks, calibration, or forecasting methodology.
• Experience with structured evidence frameworks — GRADE, systematic review protocols, or bespoke assessment rubrics.