Pattern is developing a cutting-edge AI drug discovery engine that combines AlphaFold/OpenFold structural predictions, generative molecular design, and reinforcement learning agents to navigate the ~10^60 possibilities in small-molecule chemical space. We are seeking a Deep Learning Lead to architect, train, and deploy machine learning models for protein–ligand structure prediction, de novo molecular generation, and multi-objective optimization. You will partner closely with our Ligand Design & Pose Prediction Lead to integrate chemical and biological expertise into the model pipeline, ensuring our AI agents produce compounds that are potent, novel, and biologically relevant. Key Responsibilities Design and implement deep learning architectures for: Pocket-conditioned molecular generation (e.g., Pocket2Mol, SE(3)-equivariant GNNs, 3D diffusion models). Protein–ligand pose prediction (DiffDock, EquiBind, custom SE(3)-transformers). Multi-objective reinforcement learning for compound optimization (potency, ADMET, novelty). Fine-tune AlphaFold/OpenFold and related structure models for project-specific targets. Integrate multi-modal biological context (from the Agentix Knowledge Graph) into generative and scoring models. Develop and maintain scoring functions for binding affinity, selectivity, ADMET, and synthetic accessibility. Implement uncertainty estimation and active learning loops to prioritize compound synthesis/testing. Collaborate with the Ligand Design Lead to: Translate medicinal chemistry insights into model constraints. Incorporate wet-lab feedback into model retraining. Co-develop workflows for real-time human–AI co-design. Maintain MLOps pipelines for dataset versioning, model deployment, and experiment tracking. Qualifications Required PhD or MSc in Computer Science, Machine Learning, Computational Chemistry, Bioinformatics, or related discipline. 2+ years of hands-on experience in deep learning model development, ideally in a scientific or molecular domain. Strong expertise in generative modeling (transformers, diffusion models, graph neural networks). Experience with 3D geometric deep learning for molecular structures (SE(3)-equivariant architectures). Proficiency in reinforcement learning (policy gradients, model-based RL, quality-diversity search). Strong programming skills in Python with frameworks like PyTorch or TensorFlow. Experience in handling molecular datasets (PDB, ChEMBL, binding affinity data). Preferred Prior work on protein structure prediction, ligand docking, or molecular property prediction. Familiarity with cheminformatics toolkits (RDKit, Open Babel). Experience in integrating wet-lab assay data into active learning loops. MLOps experience (Docker, CI/CD for ML, MLflow, DVC). Why Join Us? You’ll help define the AI engine at the heart of Pattern’s platform, working with cutting-edge molecular AI tools and direct structural chemistry input from our Ligand Design Lead. Your models will power a proprietary agentic search system designed to explore the chemical universe faster and smarter than competitors.
About the Role Pattern is building a next-generation AI-driven drug discovery platform that integrates state-of-the-art structural modeling, generative design, and reinforcement learning agents to explore the vast chemical space for novel small-molecule therapeutics. We are seeking a Ligand Design & Pose Prediction Lead to guide de novo small-molecule exploration, interpret protein–ligand binding predictions, and prioritize compounds for synthesis/testing. You will work in close partnership with a deep learning specialist to combine cutting-edge AI tools with your medicinal chemistry and structure-based design expertise. This is a strategic, non-lab role — your primary focus will be to bridge AI outputs with biological and chemical insight, ensuring the most promising designs move forward. Key Responsibilities Lead ligand pose prediction workflows using state-of-the-art AI and computational docking tools (e.g., DiffDock, EquiBind, Glide, GOLD). Evaluate protein–ligand binding interactions for fit, contact quality, and structural plausibility. Collaborate with AI/deep learning engineers to refine de novo molecular generation strategies using models such as REINVENT, Pocket2Mol, and diffusion-based 3D generators. Apply drug-likeness, ADMET, novelty, and selectivity criteria to prioritize compound candidates. Integrate binding mode insights with biological context from Pattern’s Agentix Knowledge Graph to align compounds with target mechanism-of-action. Generate clear, actionable compound selection lists for partner synthesis and in-vitro testing. Contribute to feedback loops by incorporating experimental assay data into ongoing model optimization. Present binding hypotheses, SAR rationale, and prioritization strategies to cross-functional teams. Qualifications Required PhD or Masters in Medicinal Chemistry, Chemical Biology, Computational Chemistry, or related discipline (or MSc + 3+ years of relevant experience). Proven experience in structure-based drug design and ligand pose evaluation. Strong working knowledge of protein–ligand binding principles (H-bonds, hydrophobic contacts, electrostatics, shape complementarity). Familiarity with AI/ML-based molecular design platforms. Ability to work with predicted protein structures (AlphaFold/OpenFold) and assess binding pockets. Experience applying drug-likeness rules and property-based filtering in lead prioritization. Excellent communication skills and ability to work cross-functionally with AI engineers, biologists, and medicinal chemists. Preferred Familiarity with pharmacophore modeling and pocket geometry analysis. Experience in multi-objective optimization (binding, ADMET, novelty). Exposure to reinforcement learning-driven compound optimization workflows. Why Join Us? You’ll be joining at the frontier of AI-guided drug discovery, working side-by-side with deep learning experts to build a platform capable of efficiently searching through 10^60 chemical possibilities. Your expertise will directly shape the quality and novelty of our candidate compounds, accelerating the path from target to therapy.