Deep Learning Research Engineer Intern

This is a fully remote position, open to applicants in Massachusetts.

📋 Description

• Create and enhance the fundamental probabilistic architecture of the model.

• Design input encoders for temporal, unordered tabular, and mixed-modality data.

• Develop attention factorizations that span variables, rows, and time horizons.

• Construct distributional output heads and decoding methods for coherent joint samples.

• Conduct controlled architectural studies, which include ablation tests, scaling behavior assessments, memory profiling, and throughput analysis.

• Develop scalable PyTorch implementations that accommodate larger input/output spaces, enhanced throughput, and optimized memory usage.

• Investigate how architectural design, data generation, and inference limitations influence benchmark performance and practical utility.

• Expand the synthetic data engine with more complex stochastic dynamics, constraints, dependence structures, heavy tails, and regime behavior where applicable.

• Transform research concepts into reliable implementations and credible empirical findings.

• Collaborate on research initiatives aimed at achieving state-of-the-art results while adhering to practical deployment considerations.

⛳️ Requirements

• Proficient in PyTorch with practical experience in building and training deep learning models.

• Strong grasp of Transformers, attention mechanisms, and long-context or state-space sequence architectures, including the trade-offs between quality, memory usage, and latency.

• Experience in probabilistic modeling within neural networks, involving distributional output heads, likelihood-based or proper-scoring-rule losses, mixture or flow models, or similar uncertainty-aware learning frameworks.

• Solid grounding in probability and statistics, including understanding densities and masses, inter-variable dependencies, and model calibration.

• Capability to translate architectural concepts into functional GPU-native implementations, controlled experiments, and diagnostics.

• Strong engineering practices, including writing clear code, conducting tests, ensuring reproducibility of experiments, and maintaining a disciplined evaluation process.

• Proficiency in debugging training instability, analyzing altered results, and iterating from hypotheses to empirical evidence.

• Familiarity with stochastic processes and stochastic differential equations is highly advantageous.

• Experience in designing synthetic data generators or simulation-based training curricula is also a significant plus.

• Depth of knowledge in finance, energy, commodities, or similarly quantitative domains is a strong advantage.

• Acquaintance with tabular or mixed-modality deep learning is beneficial.

• Post-doctoral-level experience or equivalent is highly desirable.

• Proficiency in PyTorch is required.

• Capability to work with GPU-native implementations is essential.

🏝️ Benefits

• Medical, Dental & Vision Insurance.

• Flexible Time Off Program.

• Paid Holidays.

• Paid Parental Leave.

• Global Employee Assistance Program (EAP).

• Competitive pay and benefits.