AI Algorithms Research Scientist-Vice President

Are you looking for an exciting opportunity to join a dynamic and growing team in a fast-paced and challenging area? This is a unique opportunity for you to work with the Global Technology Applied Research (GTAR) center at JPMorgan Chase & Co. The goal of GTAR is to design and conduct research across multiple frontier technologies, in order to enable novel discoveries and inventions, and to inform and develop next-generation solutions for the firm's clients and businesses.
As an AI Algorithms Research Scientist, Vice President, within the Global Technology Applied Research (GTAR) center at JPMorgan Chase & Co., you will advance the algorithmic foundations of modern AI — the methods that determine how efficiently large language models and agentic systems learn, reason, and run at scale. You will develop novel algorithms and establish their theoretical foundations, implement them in performant software, provide novel research solutions to problems faced by internal project teams, and contribute to JPMC's IP by pursuing necessary protections of generated IP.
Job Responsibilities
Advance the algorithmic foundations of large-scale AI and their applications to model training, inference, and agentic systems.
Develop novel algorithms that improve the accuracy, latency, and compute cost of large language model and agentic workloads.
Establish the theoretical grounding of the methods you develop, including convergence, approximation quality, and sample- and compute-efficiency.
Implement the developed algorithms in performant software and validate them at scale.
Provide novel research solutions to problems faced by internal project teams.
Work with other researchers to document your findings in scientific papers and present them at conferences.
Contribute to JPMC's IP by pursuing necessary protections of generated IP.
Required qualifications, capabilities, and skills
Ph.D. degree in computer science, mathematics, physics, electrical engineering, statistics, or related fields, with at least 2 years of experience (industry or postdoc).
Demonstrated research ability in AI/ML algorithms, optimization, or theory.
A deep foundation in optimization, probability, linear algebra, and learning theory. and experience in scientific technical writing.
Proficiency in Python, and C/C++ or CUDA for performance-critical work.
Experience developing performant codes.
Strong communication skills and the ability to present findings to a non-technical audience.
Experience in one or more of the following domains:
Efficient learning and inference (e.g., quantization and low precision, sparsity and pruning, distillation, low-rank and structured approximations, speculative decoding, KV-cache optimization)
Optimization (e.g., stochastic and second-order methods, optimizer and preconditioner design, training dynamics)
Foundations of scale (e.g., scaling laws, compute-optimal training, stability of large-model training)
Hardware–algorithm co-design (e.g., algorithms designed around accelerator memory hierarchy, bandwidth, and numerical precision)
Preferred qualifications, capabilities, and skills
Preference is given to candidates with a strong publication record (example venues include but are not limited to NeurIPS, ICML, ICLR, COLT, STOC, FOCS, ISCA, HPCA).
Experience with GPU/accelerator programming and profiling (e.g., CUDA, Triton, Nsight).
Contributions to open-source ML systems, libraries, or performance-critical code.
No prior familiarity with finance or financial use cases is required.
Preference given to candidates who include a link to their Google Scholar or Semantic Scholar profile in their resume.