Senior Manager, Data Science

Decision Science colleagues will serve as a key member of the Credit and Fraud Risk organization. We seek a thought-leader and a problem-solver who can blend business, technical, and industry best practices when it comes to developing the analyses, models, and algorithms that power our customers’ digital experiences. This critical team is responsible for managing enterprise risks throughout the customer lifecycle, across our consumer and commercial businesses, and across all our global products. We develop industry-first data capabilities, build profitable decision-making frameworks, create machine learning-powered predictive models, and improve customer servicing strategies. Our Decision Science teams use industry leading modeling and AI practices to predict customer behavior. We develop, deploy and validate predictive models and support the use of models in economic logic to enable profitable decisions across credit, fraud, marketing and servicing optimization engines. • Own the design, implementation, and optimization of sophisticated machine learning algorithms in modern C++, working directly from mathematical and algorithmic formulations through production implementation. • Translate mathematical and statistical concepts into efficient algorithms and production-quality C++ implementations; evaluate alternative formulations and computational approaches where appropriate. • Develop a deep understanding of existing algorithms and improve their mathematical formulation, computational design, data structures, numerical behavior, efficiency, and scalability while preserving correctness. • Profile and optimize CPU implementations using algorithmic improvements, parallelism, multithreading, vectorization, memory/cache optimization, and other high-performance computing techniques. • Design and implement CUDA C/C++ solutions to accelerate computationally intensive ML algorithms on NVIDIA GPUs, including redesigning algorithms where necessary for massively parallel execution. • Optimize GPU computation, memory access, synchronization, data movement, and hardware utilization. • Architect single- and multi-GPU execution, addressing workload decomposition, GPU communication, synchronization, memory management, and scalability. • Build and optimize GPU-enabled workloads on Google Cloud Platform (GCP). • Establish rigorous validation, numerical correctness, CPU/GPU equivalence, benchmarking, and performance-regression methodologies. • Integrate high-performance C++/CUDA implementations with Python-based ML environments and collaborate well with partner teams and downstream users. • PhDs in a quantitative field (Computer Science, Computer Engineering, Electrical Engineering, Mathematics, Physics, Statistics, and etc.) with hands-on experience developing sophisticated machine learning algorithms and techniques. Contribution to open-source project in C++/CUDA is a significant plus. • Strong foundation in mathematics, statistics, optimization, and machine learning, with the ability to understand sophisticated mathematical formulations and translate them into computational algorithms. • Deep expertise in modern C++, algorithms, and data structures, with demonstrated experience implementing and optimizing complex mathematical, numerical, or machine learning algorithms. • Ability to work across mathematical formulation, algorithm design, and software implementation, and to make performance improvements while maintaining numerical and algorithmic correctness. • Strong understanding of computational complexity and performance engineering, including memory management and locality, multithreading, concurrency, vectorization, profiling, and benchmarking. • Strong hands-on experience with CUDA C/C++ and NVIDIA GPUs, including development, debugging, profiling, and optimization of GPU software. • Strong understanding of CPU/GPU parallel algorithm design, including memory hierarchy, data movement, synchronization, workload decomposition, and efficient mapping of algorithms onto GPU hardware. • Experience with multi-GPU computing on the cloud, including workload distribution, communication, synchronization, memory management, and scaling computational workloads across GPUs. • Strong proficiency in Python and experience integrating high-performance C++/CUDA components into production machine learning or computational systems • Expertise in an analytical language (Python, R or the equivalent), and experience with databases (Hive, SQL, or the equivalent). Experience with data visualization is a plus. • Demonstrated ability to frame business problems into mathematical programming problems, leverage external thinking and tools (from academia and/or other industries) to engineer a solution and deliver business insights. • Ability to work effectively in a team environment • Independent thinker who’s organized, has great attention to detail, and can multi-task • Strong communication skills • Strong team player with a demonstrated ability to develop team members and create highly effective and results-driven culture • Strong relationship management and proven track record of positively collaborating and partnering with stakeholders • Ability to learn quickly and work independently with sophisticated, unstructured initiatives • Ability to integrate with cross-functional business partners worldwide • Proficient in presentation tools, including Excel and PowerPoint