Deep Learning Research Engineer Intern
Remote • Boston, Massachusetts, US
Posted 6d ago
Job Location
Boston, Massachusetts, US
Tech Stack
Remote Work Policy
Fully remote
Categories
AI Research Engineer
About the job
DataRobot is seeking a Deep Learning Research Engineer Intern to contribute to the development of a multi-target joint probabilistic foundation model. This ambitious project aims to create a single model capable of handling temporal forecasting, unordered tabular regression and classification, missing-data completion, and mixed-modality inputs and outputs, learning coherent joint structures across various dimensions. The primary focus is on generating real-world value, though strong benchmark performance is also expected. This internship offers a unique opportunity to shape the architecture of a probabilistic foundation model, working on encoders, attention mechanisms, sequence architectures, and distributional output heads. The role requires a blend of deep learning expertise, strong engineering habits, and the ability to reason precisely about model capabilities and limitations.
Responsibilities
- Design and improve the core model architecture, including input encoders for temporal, tabular, and mixed-modality data, attention mechanisms, and distributional output heads.
- Conduct controlled architecture studies, including ablations and scaling behavior analysis, to inform design decisions.
- Build scalable PyTorch implementations to support larger input/output spaces, improve throughput, and optimize memory usage.
- Study the interaction between architectural choices, data generation, and inference constraints to enhance benchmark quality and real-world usefulness.
- Extend the synthetic-data engine with richer stochastic dynamics and dependence structures (for candidates with relevant background).
- Translate research ideas into robust implementations and credible empirical results.
Requirements
- Strong PyTorch skills and hands-on experience building and training deep models.
- Solid understanding of Transformers, attention variants, and long-context or state-space sequence architectures, including their tradeoffs.
- Experience with probabilistic modeling in neural networks, such as distributional output heads, likelihood-based losses, or uncertainty-aware learning.
- Strong foundation in probability and statistics.
- Ability to connect architectural ideas to GPU-native implementations, controlled experiments, and diagnostics.
- Strong engineering habits, including readable code, tests, reproducible experiments, and disciplined evaluation.
- Ability to debug training instability and iterate quickly from hypothesis to evidence.