Computational Scientist

Sharp Decisions; San Francisco (CA); On-site; USD 83,000 - 138,000; Full time

Job summary

Sharp Decisions seeks a highly skilled researcher to work on multimodal perturbation data in San Francisco. The contractor will train and evaluate deep learning models in PyTorch, conduct controlled train/test experiments on a SLURM cluster, and assess how data portions affect learning outcomes.

Applicants should hold a PhD or equivalent research experience, with strong statistics and scientific Python (NumPy, pandas, scanpy/anndata) skills.

Qualifications

• PhD in a computational discipline, or equivalent research experience.
• Proficiency with PyTorch for training and evaluating deep learning models.
• Strong applied statistics and Python proficiency (NumPy, pandas, scanpy/anndata).
• Experience designing controlled train/test splits on a SLURM cluster.

Responsibilities

• Train and evaluate models on the multimodal perturbation dataset.
• Characterize which data portions contribute most to model learning.
• Evaluate model performance and model outputs against established baselines.

Skills

DL with PyTorch

Applied statistics

Python (NumPy/Pandas/Scanpy)

Education

PhD or equivalent

Tools

SLURM cluster

Job description

Duration : 12 Months Contract with a possible extension

Key Responsibilities

Multimodal perturbation datasets are becoming increasingly large and available, and understanding which data contribute most to model performance is an increasingly important question. The Regev lab has generated a large multimodal perturbation dataset. The contractor will train and evaluate models on this data to characterise how different portions of the dataset contribute to model learning, and will evaluate model performance and model outputs against

Required Education:

- PhD in a computational discipline, or equivalent research experience

Preferred majors: Computational Biology, Computational Sciences, Statistics or Machine Learning

Required skills:

• Training and evaluating deep learning models in PyTorch, including designing and executing controlled train/test split experiments on a SLURM cluster.
• Strong applied statistics
• Proficient scientific Python (numpy, pandas, scanpy/anndata) and reproducible, documented analysis practice.

Preferred Qualifications:

• Experience with single-cell transcriptomics
• Experience with analysis of CRISPR perturbation screens (perturb-seq or similar).
• Familiarity with representation learning