Machine Learning Researcher – Agentic Science

This is a fully remote position, open to applicants in United States.

📋 Description

• Lead the development of the agentic research vertical at PostEra by utilizing agentic systems to streamline the creation of mechanistic models for biochemical and physiological processes.

• Analyze biological data to validate targets and leverage models to inform drug discovery choices.

• Create machine learning techniques that adapt to new drug discovery challenges with limited labeled data.

• Construct molecular and tabular in-context learning systems and foundational models from proprietary multimodal datasets.

• Identify relevant prior examples and tasks, assess when transfer learning is beneficial or detrimental, and deliver dependable predictions amidst distribution shifts.

• Define tasks, develop datasets and evaluation scenarios, establish baselines, train and scale models, conduct thorough ablation studies, and translate successful methodologies into practical tools for scientists.

• Develop and evaluate agentic systems for quantitative models in biological and physiological contexts.

• Proactively identify, design, and oversee research initiatives focused on in-context learning, agentic systems, few-shot adaptations, tabular foundational models, and molecular machine learning.

• Design and train models that adjust to new assays, endpoints, targets, or chemical series.

• Rigorously compare methods against strong baselines and curate test cases sensitive to biases.

• Collaborate with scientists on potency modeling, ADME prediction, selectivity, lead optimization, and the design of early clinical studies.

• Develop efficient training and data pipelines while scaling models across molecular and tabular tasks.

• Generate clear, reproducible research code, maintain experiment tracking, contribute to code reviews, documentation, and shared modeling infrastructure.

• Publish findings in prominent machine learning, medicinal chemistry, or computational biology outlets, and represent PostEra within the scientific community.

⛳️ Requirements

• PhD in machine learning or a STEM field focused on developing innovative machine learning techniques.

• Proven track record of high-quality research, including publications and contributions to open-source projects.

• Strong background in research or engineering related to modern machine learning, deep learning, or statistical modeling, with a solid grasp of the theoretical foundations of machine learning algorithms.

• Demonstrated proficiency in at least one relevant area: agentic workflows for scientific research, machine learning methods for bioinformatics and clinical data modeling, in-context learning, tabular learning, or few-shot learning.

• Practical experience in training, debugging, and assessing ML models in Python with frameworks like PyTorch or JAX.

• Capable of independently translating vague scientific or technical challenges into well-defined ML projects, including datasets, task definitions, baselines, metrics, and validation strategies.

• Skillful in designing meticulous experiments, benchmarks, and ablations that differentiate genuine improvements from biases, while understanding which model aspects contributed to those improvements.

• Comfortable operating in a startup environment where priorities may shift, data may be imperfect, and high-quality judgment is as crucial as model complexity.

• Prior experience in drug discovery is not mandatory.

• Nice-to-have: experience with training tabular foundational models, particularly in sparse, heterogeneous, small-data, or high-missingness scenarios.

• Nice-to-have: experience in developing molecular in-context learning systems or adapting general-purpose in-context models to molecular or scientific data.

• Nice-to-have: experience in large model training, including models with over 1 billion parameters, distributed training, sharding, data parallelism, model parallelism, and large-scale data pipelines.

• Nice-to-have: involvement in the development of AI "co-scientist" systems for physical or biological challenges.

• Nice-to-have: hands-on experience in modeling biological, biochemical, or clinical data using machine learning techniques.

• Nice-to-have: experience in transitioning research models into production-level scientific software or computational workflows.

🏝️ Benefits

• Equity: 0.05 - 0.1%

• Proportional compensation

• Internal recognition

• Meaningful promotions