Research Scientist, AI Research

Takeda•Boston, MA

1d•$116,000 - $182,270•Onsite

About The Position

The global computational chemistry team at Takeda is seeking an exceptional scientist to help shape the future of AI-enabled drug discovery at Takeda, with a focus on induced-proximity therapeutics and structure-guided small-molecule design. Working across AI/ML, structural biology, and medicinal chemistry, you will develop cutting-edge computational approaches to explore chemical space more effectively and translate scientific advances into life-saving therapeutic impact.

Requirements

PhD in Computer/Data Science, Computational Chemistry/Biology, Chemical Engineering, Biophysics, or related field.
Demonstrated scientific originality and impact, evidenced by first-author publications in high-impact journals or top-tier ML conferences.
Hands-on with diffusion models, GNNs/CNNs, transformers, and 3D/structure-aware modeling and finetuning.
Hands-on experience with Schrödinger (Maestro, LiveDesign) and scripting to automate their use, ideally in a High-Performance Computing (HPC) environment.

Nice To Haves

Develop and benchmark structure-aware generative models for pocket-conditioned design across synthesizable, billion-scale chemical libraries.
Experience with parameter-efficient fine-tuning (LoRA / low-rank adapters) of biomolecular foundation models.
Experience with water/metal site prediction to improve docking/scoring is a plus.
Experience deploying multi-property generative AI models as containerized REST services (FastAPI, Docker) covering ADMET properties (Lipinski, BBB, CYP, hERG) for medicinal chemistry workflows.
PhD-level background in all-atom molecular dynamics simulations (LAMMPS, GROMACS, OpenMM) with automated high-throughput workflow management.
Direct collaboration with medicinal chemistry teams in a pharmaceutical R&D setting, including delivery of ML predictions into production decision-support workflows.
TPD domain exposure (e.g., CRBN glues, E3 recruitment) and ternary complex or PPI modeling.

Responsibilities

Train/finetune diffusion, transformer, and equivariant GNN models for pocket-aware ligand design, pose/affinity prediction, and sequence/structure tasks (co-folding, multimer/ternary modeling).
Build end-to-end, reproducible pipelines for MSA/co-folding feature generation, data QC, training, and evaluation with experiment tracking; accelerate on multi-GPU/HPC and cloud.
Automate large-scale ligand/mutation screening with geometry/covalency checks, pose-quality assessment, and multi-model consensus ranking.
Integrate docking/MD/MM/GBSA or related physics-based components with learned models to improve prioritization for peptides and small molecules.
Curate complex structural and proteomics datasets; design retrospective/prospective benchmarks and report robust metrics.
Package models as services/containers; deliver intuitive analysis apps/dashboards (e.g., Streamlit) and visualizations (e.g., PyMOL integrations).
Partner with medicinal chemists, structural biologists, and data scientists; plan experimental validation and communicate decisions clearly.
Contribute to internal methods notes; author publications/posters for leading ML and scientific venues.