Data Engineer

Your impact

Engineer the scientific data foundations that determine what our models can learn and how confidently our team can evaluate them.

As a Data Engineer, you will turn protein sequences, structures, annotations, and experimental results into reliable, reproducible datasets for model training, evaluation, and molecular design. The quality of this work directly affects model behavior, which molecules scientists prioritize, and what the team learns from experiments.

This is a hands-on engineering role. You will own data meaning, quality, lineage, and delivery while working closely with AI researchers and computational biologists to define fitness for scientific use and with software engineers on the storage and execution systems that support it.

What you will do

• Build scientific data pipelines. Ingest, normalize, and integrate public, internally generated, and experimental datasets while handling changing formats and schemas.
• Curate data for learning and discovery. Implement filtering, clustering, deduplication, and feature generation with researchers, making selection criteria explicit and reproducible.
• Make data quality measurable. Define contracts and automated checks for completeness, consistency, and scientific validity, and trace downstream problems back to their source.
• Support trustworthy model evaluation. Construct training, validation, and test sets; control biological overlap; investigate contamination and leakage; and document benchmark assumptions.
• Own dataset releases and provenance. Version sources, transformations, and outputs so releases are auditable, comparable, and reproducible.
• Make scientific data efficient and accessible. Improve data representation, processing, and delivery, and build documentation and exploration tools that help scientists understand coverage, quality, and limitations.

What you bring

You are an engineer who cares about what data means, not only whether a pipeline completes. You can translate evolving research questions into explicit contracts and transformations, investigate subtle problems systematically, and explain the evidence behind a dataset.

• Experience building and maintaining data pipelines or data-intensive software used by others.
• Strong Python skills and an understanding of data modeling, schemas, storage formats, and efficient processing.
• A systematic approach to validation and debugging, including investigating silent errors and tracing downstream results to their source.
• Experience processing data beyond a single machine, or clear evidence that you can reason about partitioning, parallelism, memory limits, and recoverable execution.
• The ability to turn research questions into concrete data requirements, explain trade-offs, and document assumptions clearly.
• Ownership, curiosity, and care for detail across ingestion, transformation, release, and use.

You do not need experience with every tool in our stack. We value strong engineering fundamentals and the ability to learn the scientific meaning of the data you work with.

Experience that would help

• Protein sequence and structure data, molecular datasets, or another scientific domain where representation and provenance affect interpretation.
• Dataset construction for machine learning, including sampling, evaluation splits, leakage analysis, and distribution shifts.
• Scientific Python, ML data tooling, and bioinformatics programs such as NumPy, PyTorch, and MMseqs2.

Our working environment

Our stack includes Python, Rust, and TypeScript, with Docker and AWS alongside on-premises GPU infrastructure. You will work primarily on scientific data processing and delivery, collaborating with software engineers on execution and storage and with researchers on curation and evaluation.

AI coding tools help us explore formats, implement transformations, develop checks, and automate repetitive work. We remain accountable for explaining what changed, what was preserved, and why the resulting data is suitable for its scientific purpose.

Why Pythia

The datasets you build will shape model quality, the evidence researchers trust, and the molecules selected for experimental testing. In a small team, you can follow that impact from raw data through model development and scientific decisions.

We are a small, interdisciplinary team. We value people who take ownership, move between disciplines, check assumptions against evidence, and bring others into decisions. You will have room to shape both the work and how the team works.