Advisor - Antibody Developability Validation & Benchmarking
Eli Lilly and Company
3 days ago
On-site
Indianapolis, IN
Clinical Research and Development
Purpose
Advisor/Senior Advisor – Antibody Developability Validation & Benchmarking for Lilly TuneLab (AI-powered federated learning platform). Establish trustworthiness of federated antibody models for triage/go-no-go decisions; contribute to model design with close partnership on architecture, feature design, and uncertainty quantification.
Key Responsibilities
- Build canonical developability benchmark suite (aggregation propensity, thermal stability, polyspecificity, self-interaction, viscosity, chemical liabilities, immunogenicity surrogates) and define endpoint evaluation/roll-up for triage.
- Design sequence-aware federated test sets with privacy-preserving, representative splits (germline-, CDR-similarity-, clonotype-based) and held-out evaluation accounting for antibody data asymmetry.
- Integrate and benchmark against external public resources (SAbDab, OAS, TAP, Jain panel, FLAb, emerging datasets) to quantify generalization gaps vs public-only baselines.
- Develop cross-domain validation across IgG/bispecific/fragment formats, expression systems, and partner assay protocols.
- Implement temporal and sequence-similarity-aware validation to simulate deployment, detect concept drift, and surface failure modes.
- Partner on validation-impacting model design choices (uncertainty quantification, calibration, structure-aware vs sequence-only, endpoint combination/independence).
- Ensure statistical rigor (multiple testing, hierarchical structures, confidence intervals).
- Build reproducibility MLOps infrastructure (data snapshot/model checkpoint/hyperparameter versioning).
- Perform performance profiling across germline/length regimes/property ranges.
- Integrate validation frameworks with TuneLab (NVIDIA FLARE) for scalable automated testing.
Basic Qualifications
- PhD in Computational Biology/Bioinformatics/Computational Chemistry/Computer Science/Statistics or related.
- 4+ years post-PhD in antibody discovery/engineering/developability data in biopharma or academic setting.
- Experience analyzing/modeling antibody developability assay data (e.g., HIC, AC-SINS, nanoDSF, polyspecificity panels, viscosity, chemical liabilities).
- Hands-on antibody numbering tools (ANARCI or equivalent) and knowledge of Kabat/Chothia/IMGT.
- Experience designing ML validation protocols for biological sequence data (similarity-aware splits, held-out test design).
Additional Preferences
- Fine-tune protein/antibody language models; knowledge of sequence liability motifs.
- Experimental design, statistical validation, hypothesis testing.
- Data engineering, automation; NVIDIA FLARE or similar federated frameworks.
- Antibody structure prediction tools (AlphaFold-Multimer, IgFold, ABodyBuilder).
- Familiarity with SAbDab/OAS/TAP/Jain/FLAb; manufacturability funnel.
- Regulatory considerations for AI/ML; uncertainty quantification + calibration.
- PyTorch, ML ecosystem (HF, scikit-learn, RDKit), MLflow/W&B; publications; scientific rigor; strong technical writing; partner-facing model cards/reports.
Location/Travel
Indianapolis, San Francisco, or Boston; up to 10% travel.
Compensation & Benefits (if included)
- Anticipated wage: $166,500–$266,200; eligible for company bonus and comprehensive benefits (401(k), medical/dental/vision, life insurance, time off/leave, well-being).
Advisor/Senior Advisor – Antibody Developability Validation & Benchmarking for Lilly TuneLab (AI-powered federated learning platform). Establish trustworthiness of federated antibody models for triage/go-no-go decisions; contribute to model design with close partnership on architecture, feature design, and uncertainty quantification.
Key Responsibilities
- Build canonical developability benchmark suite (aggregation propensity, thermal stability, polyspecificity, self-interaction, viscosity, chemical liabilities, immunogenicity surrogates) and define endpoint evaluation/roll-up for triage.
- Design sequence-aware federated test sets with privacy-preserving, representative splits (germline-, CDR-similarity-, clonotype-based) and held-out evaluation accounting for antibody data asymmetry.
- Integrate and benchmark against external public resources (SAbDab, OAS, TAP, Jain panel, FLAb, emerging datasets) to quantify generalization gaps vs public-only baselines.
- Develop cross-domain validation across IgG/bispecific/fragment formats, expression systems, and partner assay protocols.
- Implement temporal and sequence-similarity-aware validation to simulate deployment, detect concept drift, and surface failure modes.
- Partner on validation-impacting model design choices (uncertainty quantification, calibration, structure-aware vs sequence-only, endpoint combination/independence).
- Ensure statistical rigor (multiple testing, hierarchical structures, confidence intervals).
- Build reproducibility MLOps infrastructure (data snapshot/model checkpoint/hyperparameter versioning).
- Perform performance profiling across germline/length regimes/property ranges.
- Integrate validation frameworks with TuneLab (NVIDIA FLARE) for scalable automated testing.
Basic Qualifications
- PhD in Computational Biology/Bioinformatics/Computational Chemistry/Computer Science/Statistics or related.
- 4+ years post-PhD in antibody discovery/engineering/developability data in biopharma or academic setting.
- Experience analyzing/modeling antibody developability assay data (e.g., HIC, AC-SINS, nanoDSF, polyspecificity panels, viscosity, chemical liabilities).
- Hands-on antibody numbering tools (ANARCI or equivalent) and knowledge of Kabat/Chothia/IMGT.
- Experience designing ML validation protocols for biological sequence data (similarity-aware splits, held-out test design).
Additional Preferences
- Fine-tune protein/antibody language models; knowledge of sequence liability motifs.
- Experimental design, statistical validation, hypothesis testing.
- Data engineering, automation; NVIDIA FLARE or similar federated frameworks.
- Antibody structure prediction tools (AlphaFold-Multimer, IgFold, ABodyBuilder).
- Familiarity with SAbDab/OAS/TAP/Jain/FLAb; manufacturability funnel.
- Regulatory considerations for AI/ML; uncertainty quantification + calibration.
- PyTorch, ML ecosystem (HF, scikit-learn, RDKit), MLflow/W&B; publications; scientific rigor; strong technical writing; partner-facing model cards/reports.
Location/Travel
Indianapolis, San Francisco, or Boston; up to 10% travel.
Compensation & Benefits (if included)
- Anticipated wage: $166,500–$266,200; eligible for company bonus and comprehensive benefits (401(k), medical/dental/vision, life insurance, time off/leave, well-being).