About the role#
This is a research-engineering internship for individuals operating at or near the post-doctoral level. You will shape the architecture of a multi-target joint probabilistic foundation model designed to solve complex real-world problems. The project focuses on building a system capable of temporal forecasting, unordered tabular regression, classification, and missing-data completion while maintaining coherent joint structure across variables and scenarios. We prioritize creating real value in serious use cases over isolated benchmark optimization.
What you'll do#
- Design and improve core model architecture, including input encoders for temporal and tabular data, attention mechanisms, and distributional output heads.
- Implement, train, and evaluate PyTorch models within a production-quality codebase.
- Run controlled architecture studies, such as ablations, scaling behavior, and throughput profiling, to inform design decisions.
- Build scalable PyTorch implementations that optimize for memory and throughput.
- Extend the synthetic-data engine with stochastic dynamics, constraints, and dependence structures.
- Translate research ideas into robust implementations and credible empirical results.
What you'll need#
- Strong PyTorch skills and hands-on experience building and training deep models.
- A solid understanding of Transformers, attention variants, and sequence architectures.
- Experience with probabilistic modeling in neural networks, such as distributional output heads or likelihood-based losses.
- A strong foundation in probability and statistics to reason about densities, dependence between variables, and calibration.
- Excellent engineering habits, including readable code, tests, and reproducible experiments.
- The ability to debug training instability and iterate quickly from hypothesis to evidence.
- Academic or professional background in Computer Science, Mathematics, or Statistics.
- Working knowledge of stochastic processes or differential equations, experience with synthetic data generators, or familiarity with tabular/mixed-modality deep learning are considered pluses.
Location & details#
- This is a full-time, paid internship.
- The role is available on-site in Boston, Massachusetts, or San Francisco, California, as well as remotely within the United States.
- We accept applications on a rolling basis.
About DataRobot
DataRobot provides an agent workforce platform for enterprise AI development. The company offers tools for model development, MLOps, and governance. It supports AI experimentation and deployment across various cloud environments. Founded in 2012, the firm is based in Boston and employs roughly 880 people.
How to get in at DataRobot
Applying early gives you a distinct advantage at DataRobot, as recruiters often review candidates as they arrive. Intern Insider sends an instant alert the moment a role matching your target is published anywhere, so you apply among the first before the pile grows. This helps you avoid the common frustration of submitting an application to a role that has already been filled. You can also improve your chances by reaching out to the right people. Intern Insider surfaces the recruiters behind the company roles so you can reach out directly to ask about the position or a referral. Connecting with a recruiter directly often improves your response rates compared to waiting for a system update.
