Senior ML Research Engineer, Virtual Cell

Entangled Future; Palo Alto (CA); Hybrid; USD 180,000 - 240,000; Full time

Benefits offered by this job

Competitive base salary

Equity participation

Comprehensive benefits

Job summary

SandboxAQ is seeking a Senior ML Research Engineer for the Virtual Cell project in the BioSim team, focusing on building models and data infrastructure for AQCell. You will handle diverse transcriptomic and endpoint datasets and drive model training, evaluation, and scalability to support drug discovery decisions.

You will translate ideas from literature into robust, tested code, collaborate with biologists and engineers, and contribute to scalable workflows, baselines, and reproducible research

Qualifications

• Experience building and maintaining ML models in a scientific setting.
• Experience in life sciences is preferred.
• Experience with large-scale data ingestion, cleaning, versioning, and pipeline maintenance.

Responsibilities

• Model development: build, train, and maintain ML models for expression-perturbation prediction and downstream functional endpoints.
• Large-scale dataset management across multiple datasets and harmonization efforts.
• Evaluation & baselines: automate evaluation and implement robust baselines.
• Translate scientific ideas into production-ready code within the modeling framework.
• Collaborate with biologists, software engineers, and product stakeholders to ensure biology-grounded models.
• Document methods, assumptions, and results clearly for stakeholders.

Skills

Python

ML Frameworks

Experiment Tracking

Data Management

Education

Bachelor's degree in a scientific field

MS or PhD preferred

Tools

PyTorch

JAX

Weights & Biases

Job description

Senior ML Research Engineer, Virtual Cell at SandboxAQ, United States (remote). A senior-level full time position in the BioSim team, listed under software engineering.

The Opportunity

The AI Sim R&D team builds leading-edge ML and physics-based models ("LQMs") to advance drug discovery. Within this team, AQCell is our virtual cell platform: it takes a cell representation (e.g. basal gene expression) and a perturbation descriptor (e.g. SMILES, dose, and time), predicts the resulting transcriptomic response, and maps that response through pathway activity to functional endpoints such as cell viability, IC50, and toxicity dose-response — helping drug discovery scientists understand the biological repercussions of a compound across the cell, not just whether it binds its target. As a Machine Learning Engineer on AQCell, you will build and maintain the models and data infrastructure that power this pipeline. You will work across large, heterogeneous transcriptomic and functional-endpoint datasets (e.g. LINCS L1000, GDSC, Tahoe-100M, and DILImap), train and evaluate expression-perturbation and cell-viability prediction models over them, and help harden our evaluation pipeline and baselines so we can trust and improve model performance over time. This role sits at the intersection of machine learning and biology, and offers the opportunity to shape how virtual cell models are trained, validated, and scaled toward real drug discovery decisions.

Why Join Us? We offer competitive compensation, a comprehensive benefits package, and opportunities for professional growth.

• Compensation: Competitive base salary, performance-based incentives or bonuses (where applicable), and equity participation.
• Benefits: Comprehensive medical, dental, and vision coverage for employees and dependents with generous employer premium contributions, retirement savings with company matching, paid parental leave, and inclusive family-building benefits.
• Work-Life Balance: Flexible paid time off, company-wide seasonal breaks, and support for flexible work arrangements that enable sustainable performance.

Responsibilities

• Model Development: Build, train, and maintain machine learning models for expression-perturbation prediction (e.g. transcriptomic response to a drug perturbation) and downstream functional-endpoint prediction (e.g. cell viability, IC50, toxicity dose-response).
• Large-Scale Dataset Management: Acquire, harmonize, and manage large-scale biological datasets (e.g. LINCS L1000, GDSC, Tahoe-100M, DILImap) — including schema harmonization, normalization, and de-duplication across cell, drug, and assay identifiers — and manage model training pipelines over these pooled datasets.
• Evaluation & Baselines: Contribute to automating and hardening the end-to-end evaluation pipeline, including implementing robust statistical baselines (e.g. cell- and drug-conditioned mean baselines) to rigorously benchmark model performance.
• Research Translation: Translate ideas from the scientific literature (e.g. transformer-based perturbation models, knowledge graph and GNN-based embeddings) into working, well-tested code integrated into our modeling framework.
• Cross-Functional Collaboration: Partner with computational biologists, software engineers, and product stakeholders to ensure models are grounded in sound biology and are usable in real drug discovery workflows.
• Communication: Clearly document methods, assumptions, and results, and communicate findings to both technical and non-technical stakeholders.

Requirements

• Academic Foundation: Bachelor's degree in a scientific or quantitative field (Computer Science, Physics, Mathematics, Biology, Chemistry, or related); an advanced degree (MS or PhD) is preferred.
• Applied ML in Science: Demonstrated experience building and maintaining machine learning models in a scientific discipline in an industry setting, including taking models from prototype through validation and maintenance; experience in a life sciences setting is preferred.
• Large-Scale Data Management: Required experience managing large-scale datasets and managing the training of models over them, including data ingestion, cleaning, versioning, and pipeline maintenance at scale.
• Software Engineering: Strong Python programming skills and experience with modern ML frameworks (e.g. PyTorch, JAX) and experiment tracking/data versioning tools (e.g. Weights & Biases).
• Scientific Rigor: Ability to design sound evaluation methodology (e.g. train/test splitting strategies, held-out generalization tests) and to critically interpret model performance against meaningful baselines.
• Bioinformatics & Computational Biology: Experience with bioinformatics and computational biology data analysis tools, particularly transcriptomics harmonization and normalization tools (e.g. batch correction, pseudobulking, gene ID mapping/standardization).
• Domain Datasets: Familiarity with public perturbation or drug-sensitivity datasets such as LINCS L1000, GDSC, DepMap, or single-cell perturbation atlases.
• Knowledge Graphs & GNNs: Experience with knowledge graph embeddings or graph neural networks applied to drugs, targets, or cells.
• Cheminformatics: Familiarity with cheminformatics representations (SMILES, InChI keys, PubChem/Cellosaurus identifiers) used to harmonize drug and cell metadata across datasets.
• Collaborative Innovation: Experience working in interdisciplinary environments where AI intersects with the biological sciences.