Research Scientist, Foundational Data Science

About Prior Labs

Prior Labs is focused on a major gap in AI: while foundation models have reshaped text and images, structured data has remained largely underserved. The company is building tabular foundation models to handle the data format that powers clinical trials, financial modeling, scientific experimentation, and business decision-making.

The team has already made major progress in tabular foundation models, with TabPFN v2 recognized as a Nature cover story and establishing a new benchmark for tabular machine learning. Since launch, the model has grown substantially in capability, reached more than 3.5 million downloads, and earned over 7,500 GitHub stars. Adoption is expanding across both research and industry use cases, including healthcare, transportation, and clinical decision support.

The next stage of work is even more ambitious: scaling tabular foundation models to millions of rows, thousands of features, real-time inference, and additional data modalities, while also building the production infrastructure needed for demanding real-world environments.

The organization is a selective team of 30+ people across engineering, research, and go-to-market functions, with experience from leading institutions and companies such as Google, Apple, Amazon, DeepMind, Meta, Microsoft Research, G-Research, Jane Street, Goldman Sachs, and CERN. It is led by Frank Hutter, Noah Hollmann, and Sauraj Gambhir, with guidance from notable researchers including Bernhard Schölkopf and Yann LeCun.

In 2025, the company raised €9m in pre-seed funding led by Balderton Capital and backed by leaders from Hugging Face, DeepMind, and Black Forest Labs.

Role overview

This position sits at the center of foundational data science work: advancing the building blocks of tabular foundation models so one model can address a broad range of data science problems. About half the role involves creating new frontier tools for these models, and the other half is focused on assembling the dataset and benchmark foundations that support them.

What you will do

• Design and develop new frontier capabilities for TabPFN, including reasoning, scaling, and agentic behavior, as well as methods that help one model generalize across the full spectrum of data science tasks.
• Help define the research agenda by selecting which capabilities and benchmarks matter most, focusing on meaningful problems rather than simply chasing externally defined scores.
• Incorporate external research insights and real customer needs into new model and tooling directions, and produce leading-edge results that advance the field.
• Create robust benchmarks built from structured data tied to important real-world problems, so model evaluation reflects practical performance instead of a single leaderboard.
• Implement baseline methods and competing models with high fidelity so the team can compare TabPFN accurately against applied data science standards.
• Develop an automated, agent-assisted pipeline with human oversight to scale dataset and benchmark creation while preserving rigor.

What we are looking for

• Experience solving data science problems across multiple domains and datasets with consistently strong results over a broad set of tasks.
• Comfort using a wide ML toolkit, including strong performance with gradient-boosted trees such as XGBoost, not just deep learning methods.
• Solid understanding of common data quality and benchmark issues such as leakage, label noise, distribution shift, duplication, and mislabeled targets.
• Interest in foundational, long-horizon work and willingness to take on difficult problems that others may have avoided.
• Ability to operate as a senior individual contributor in a fast-moving, early-stage, low-process setting, with strong judgment and clear opinions on data science best practices.

Nice to have

• Background building or extending evaluation systems, benchmark collections, or experiment frameworks used by others.
• Experience creating LLM- or agent-assisted workflows with human review to scale manual processes.
• Experience connecting external research or customer requirements to an internal roadmap for a model or product.
• Previous work in tabular data, structured data, foundation models, or contributing to an emerging research area through community involvement.

Life at Prior Labs

The company operates as a compact, ambitious team tackling a difficult AI problem. The environment emphasizes strong craftsmanship, rigorous thinking, and collaboration with people who care about both impact and quality.

The team values speed, depth, and doing things carefully. It is a strong fit for people who enjoy hard technical problems, want real-world impact, and are excited to help build something meaningful.

Most roles are based in Berlin, Freiburg, or New York, and the company generally expects in-office work because close collaboration matters for the complexity of the mission. Exceptional remote arrangements may be possible and typically involve frequent travel to one of the offices, along with regular company-wide offsites.

Commitments and inclusion

The company welcomes applicants from all backgrounds and identities, especially people who may not meet every listed criterion but still bring strong potential.

It is committed to a safe, inclusive workplace and equal opportunity regardless of gender, sexual orientation, origin, disability, or other personal characteristics.

Privacy and data handling

The company notes that applicant data is handled carefully and refers candidates to its recruiting privacy notice for details on what is collected, why it is collected, and how long it is retained.