QuantitativeMLEngineer(PyTorch & PPNR Migration)

Job Title: Quantitative ML Engineer (PyTorch & PPNR Migration)

Location: New York (Hybrid)


Role Objective
We are looking for a Quantitative ML Engineer to lead the technical migration of complex PPNR (Pre-Provision Net Revenue) forecasting models from a Hadoop/C++/R environment to a modern Databricks and PyTorch ecosystem. You will be responsible for translating legacy mathematical logic into optimized PyTorch tensors while ensuring strict numerical parity required for US regulatory compliance (CCAR/DFAST).
Key Responsibilities
Model Translation: Reverse-engineer legacy C++ and R codebases to extract core mathematical logic, econometric formulas, and simulation parameters.
PyTorch Implementation: Re-implement these models in PyTorch, utilizing advanced features like torch.nn for modularity and custom Autograd functions where necessary.
Optimization: Refactor code to leverage Databricks' distributed computing and PyTorch's GPU/parallel processing capabilities to reduce model execution time.
Data Integration: Build high-performance pipelines from Snowflake into Databricks using Spark and PyTorch DataLoaders.
Parity & Validation: Conduct rigorous back-testing and sensitivity analysis to ensure the new PyTorch models yield results statistically identical to the legacy Hadoop outputs.
Regulatory Documentation: Collaborating with Model Risk Management (MRM) to document the migration process, architectural changes, and validation results in compliance with SR 11-7 standards.
Required Technical Skills
Frameworks: Expert-level PyTorch (specifically for non-computer vision tasks like time-series, regression, or Monte Carlo simulations).
Languages: High proficiency in Python and a strong ability to read and interpret C++ and R (specifically statistical packages like lme4 or forecast).
Platforms: Hands-on experience with Databricks (MLflow, Spark) and Snowflake (Snowpark is a plus).
Quantitative Finance: Deep understanding of statistical modeling, econometric forecasting, or financial risk management.
Big Data: Experience migrating workloads out of Hadoop/Hive environments.
Preferred Qualifications
Experience specifically with PPNR, CCAR, or DFAST regulatory modeling.
Masters or PhD in a quantitative field (Statistics, Financial Engineering, Physics, or Math).
Experience with TorchScript or ONNX for model productionisation.