Job Title: Deep Learning Engineer | Molecular Modeling & MLOps (no c2c) Location: San Diego, CA - hybrid Industry: Pharmaceutical JOB DESCRIPTION: We are seeking a highly skilled and motivated Machine Learning Engineer to unify, extend, and benchmark state-of-the-art geometric graph and language representations for molecular property prediction. You will build upon, benchmark, and integrate the latest model architecture(s) across chemical modalities (small molecules, peptides, antibodies, etc.) while maintaining a robust, modular, and scalable codebase. Experience with scaled model training, hyperparameter tuning, tracking, and deployment utilizing tools such as Kubernetes and MLflow (or equivalent) are required. RESPONSIBILITIES: • Model Integration & Extension • Familiarity and understanding of the latest architecture improvements for language models (e.g., rotary embeddings, Low Rank Adaptors, structure-conditioning, etc.) and geometric graph neural networks (e.g., SE(3) equivariant GNNs, invariant feature GNNs, all-atom representations, etc.) for molecule representation learning • Familiarity and understanding of the latest multimodal fusion architectures, leveraging both language and structure representations • Experience integrating neural network architectures (e.g., pytorch, pytorch geometric, JAX, etc.) into a unified, modular and extensible application • Benchmarking & Evaluation • Design and execute rigorous benchmarking protocols to compare model performance on internally curated datasets • Implement automated hyperparameter tuning (e.g., via Ray Tune) and cross-validation workflows, with standardized logging and result reporting • MLOps, Deployment & Scalability • Containerize the unified codebase for reproducible execution and scalable deployment. • Develop and maintain Kubernetes workflows for distributed inference, leveraging GPU/CPU resources efficiently. • Integrate MLflow for experiment tracking, model versioning, and deployment lifecycle management. • Enable cloud/on-premises deployment, with support for model serving and monitoring in production environments. • Collaboration & Documentation • Collaborate with computational chemists, structural biologists, and data scientists to refine requirements and validate results. • Produce clear, maintainable documentation and contribute to internal knowledge transfer. REQUIREMENTS: • Advanced proficiency with Python, PyTorch, PyTorch Geometric, and deep learning libraries for molecular and protein modeling (e.g., RDKit, DeepChem, BioPython). • Demonstrated experience with both graph neural networks and transformer-based language models in molecular property prediction tasks. • Hands-on experience with Kubernetes for orchestration of ML workflows, and MLflow for experiment and model management. • Strong understanding of containerization (Docker) and cloud computing environments (AWS, GCP, or Azure). • Experience with distributed training frameworks and workflow automation. • Prefer experience with multimodal model fusion (combining graph and language representations). • Prefer prior contributions to open-source molecular ML projects. • Strong communication skills and experience working in interdisciplinary teams. About Theoris: Our goal is to Fuel Your Career! As a Theoris team member, you join a culture based on people-centered values and an environment that fosters both personal and professional growth. We build long-term relationships with our clients and our consultants. With over 30 years of building strong relationships in the industry, we’re uniquely positioned to make the right connections. This knowledge is used to find the right job placement. Our recruiting teams are experts dedicated to the information technology and engineering staffing space and are highly respected by our client base. Best-In-Class-Benefits We are in the people business; treating people right is our ONLY priority. Theoris Services consultants are full-time employees with full benefits, including: • Robust Health Insurance • 401(k) plan • PTO accrual • Paid holidays • Excellent cash-based referral program