Scientist, AI/ML — Antibody Developability

Datapoints Team | Ginkgo Bioworks | Boston, MA

About the Role

Datapoints, Ginkgo Bioworks' Bio × AI data generation platform, is seeking a Scientist to build and benchmark machine learning models that predict and design antibody developability. You will sit between our PROPHET-Ab high-throughput biophysical platform and the models it enables, working with both newly generated customer datasets and the GDPa public dataset series (clinical IgGs, sequence-diverse natural IgGs, bispecifics, VHH-Fcs), cross-format prediction, and generative design campaigns.

This computational role requires deep understanding of biophysical assay measurements, noise characteristics, and rigorous evaluation strategies for small-scale datasets. Ideal candidates are recent Ph.D. graduates with strong applied ML proficiency and interest in protein biophysics, motivated by integrated experimental and computational design.

Key Responsibilities

• Develop and validate predictive models for developability using PLM embeddings, structural features, and physicochemical descriptors.
• Implement leakage-aware evaluation frameworks for rigorous performance assessment of small-scale datasets.
• Evaluate cross-format transferability (IgG to VHH-Fc, bispecifics) and deploy data-efficient training strategies.
• Perform rigorous QC on biophysical assay data to identify outliers and maintain high data integrity for modeling.
• Maintain and develop automated pipelines to ensure reproducibility and facilitate technical reporting.
• Disseminate findings through technical reviews, peer-reviewed manuscripts, and external scientific presentations.

Qualifications

Required

• Ph.D. in Computational Biology, Bioinformatics, ML, Biophysics, or related quantitative field with emphasis on applied ML.
• Proficiency in Python and the scientific computing stack (NumPy, pandas, PyTorch/TensorFlow).
• Demonstrated experience in supervised learning on biological datasets, including expertise in cross-validation and bias mitigation.
• Knowledge of protein representations, including PLM embeddings (e.g., ESM, AbLang) and structural featurization.
• Ability to analyze biophysical data with respect to signal-to-noise ratios and experimental dynamic range.
• Effective communication across multidisciplinary teams and ability to manage concurrent technical objectives.

Preferred

• Familiarity with antibody formats (VHH, scFv), therapeutic developability liabilities, and relevant characterization assays.
• Experience with state of the art supervised techniques such as tabular foundation models.
• Exposure to generative sequence modeling, diffusion models, and reward-based steering.
• Experience with protein structure prediction (AlphaFold, ABodyBuilder) and surface patch analysis.
• Prior engagement with high-throughput experimental design and collaborative data generation.
• Record of scientific publication and software engineering fundamentals (Git, testing, cloud compute).

Location: Boston, MA (Hybrid).