Founding AI Scientist (AI × Computational Biology)

Vale Biolabs is building an automated organoid screening platform that measures how compounds perturb human biology across multiple readouts.

Our goal is to turn this rich, multi-modal experimental data into a proprietary AI pipeline that predicts compound effects and prioritizes candidates.

The role You’ll be our first dedicated computational hire and own the design and build of our data analysis and machine-learning pipeline end to end.

This is a founding role: high ownership, direct impact on scientific and product direction, and the chance to shape the team as we grow.

The ideal candidate sits at the intersection of bioinformatics and machine learning.

What you’ll do Compound representation: convert compounds into embeddings/descriptors from cheminformatics libraries for downstream modeling.

Design and train models from scratch.

You’ll architect, train, tune, and evaluate your own models rather than fine‑tuning off‑the‑shelf APIs.

In practice that means building a transformer‑based forward‑prediction model that maps a compound embedding through successive time‑step readouts to a final marker‑based outcome, and making it work on a small in‑house dataset where good architecture, regularization, and evaluation matter more than raw scale.

Build ensembles, tune hyperparameters, and set up rigorous held‑out‑compound evaluation so you can prove the model works on molecules it has never seen.

Set up environment management, versioning, and reproducibility standards (packaging, dependencies, documentation).

Work closely with our wet‑lab scientists to align data generation with modeling needs, and help recruit and mentor future computational hires.

Must have qualifications Advanced degree spanning bioinformatics/computational biology and AI/ML (e. g., Ph D bioinformatics + MSc AI, or vice versa).

Strong Python (Biopython, pandas, Num Py, Sci Py) and R (Bioconductor, ggplot2, dplyr); Bash and hands‑on pipeline/environment setup.

Solid grounding in biological data processing and core inputs (transcriptomics, imaging/phenomics, genomics/biomarkers).

Working knowledge of deep learning (representation learning, VAEs, transformers) sufficient to prototype models.

Bonus (growth areas, not blockers) VAE embeddings with ZINB/negative‑binomial loss; sc VI / sc Gen / total VI / Multi VI and sc RNA‑seq batch correction; L1000 / CMap signatures

• molecular foundation models, ADMET, SELFIES, JT‑VAE, SBDD/scaffold work
• multi‑task toxicity + SHAP, EC50–IC50 modeling, hit triaging, Mo A deconvolution
• Cell Profiler / Cell Painting, R‑CNN segmentation (a dedicated imaging/segmentation specialist is a future hire).

Who you are A builder comfortable with ambiguity and founding‑stage scrappiness.

You can stand up a pipeline from scratch and iterate fast.

This is an absolute must.

Rigorous and reproducible by instinct; you care about clean, well‑documented, testable analysis.

A clear communicator who can translate between wet‑lab biology and computational methods.

Excited to grow into technical leadership as the team scales.

What we offer Founding‑team equity.

Direct ownership of Vale’s computational strategy and platform.

A frontier problem at the intersection of organoids, multi‑modal data, and AI.

How to apply Send your CV and a short note (plus Git Hub / publications / portfolio if available) to ilaria@vale. bio.

Tell us about a data or ML pipeline you built end to end and the impact it had. #J-18808-Ljbffr