FL94 Inc. is seeking a senior machine learning research scientist to develop predictive and generative AI to accelerate small molecule lead optimization for protein editing in biotechnology.
Requirements
- 5+ years of ML drug discovery experience: Proven track record of applying machine learning to solve problems in lead optimization, such as QSAR/ADMET modeling, hit-to-lead, or active learning within a design-make-test-analyze (DMTA) cycle.
- Core ML & Python expertise: Strong proficiency in Python and modern deep learning frameworks like PyTorch/TF, with experience building and deploying production-level ML systems in a fast-paced drug discovery startup environment.
- Expertise in generative AI & Geometric Deep Learning: Demonstrated experience in developing and applying generative models (e.g., conditional flow matching, diffusion, VAEs) and graph-based or 3D (multi-task) neural networks for molecular applications.
- Proficiency with Multi-Modal Models: Experience applying multi-modal architectures to fuse molecular structures and assay data with biological context from text and other modalities, enhancing the predictive power for QSAR/ADMET properties.
- Deep domain knowledge: A solid understanding of medicinal chemistry principles (e.g., SAR, MPO) and cheminformatics toolkits (e.g., RDKit).
- A strong publication record: A track record of first-author publications in premier machine learning or computational chemistry venues (e.g., NeurIPS, ICML, J. Med. Chem., JCIM) or relevant patents.
- Open-source contributions: Meaningful contributions to major open-source projects in cheminformatics or machine learning (e.g., RDKit, DeepChem, PyG, Hugging Face).
Responsibilities
- Develop predictive ADMET/QSAR models: Design, build and/or fine-tune cutting-edge global and local models for potency, selectivity, and key ADMET properties using state-of-the-art architectures.
- Leverage publicly available foundation models (e.g., TxGemma) and data to augment sparse functional data. Fine tune internal state-of-the-art models and design objective functions for Multi-Property Optimization (MPO).
- Enable synthesis-aware design: Integrate retrosynthesis prediction and reaction modeling into the design process to ensure that generated molecules are readily synthesizable.
- Build robust ML infrastructure: Establish and maintain data pipelines, stringent benchmarks and validation frameworks for rigorous, prospective model evaluation.
- Deploy models that directly impact project decisions in our drug discovery programs.
- Design, develop and benchmark machine learning T models from public and in-house assay data.
- Design and implement active learning MPO strategies to reduce the number of compound optimization cycles in DMTA.
Other
- Reporting directly to the CTO
- Collaborate with other ML scientists, engineers and medicinal chemists
- Driven and a bias-to-action mindset: A proactive and detail-oriented approach, with excellent cross-functional communication skills for working closely with an interdisciplinary team of chemists, biologists, and engineers.
- Experience comes in many forms, skills are transferable, and passion goes a long way.
- Dedicated to building diverse and inclusive teams
