Your project will have a direct impact on JPMorgan’s businesses, will be integrated into our product pipelines, or be part of published research in top AI conferences. Full-time employment offers may be extended upon successful completion of the program within our hybrid work model. Job responsibilities Perform exploratory data analysis, feature engineering, model development, and evaluation using structured and unstructured datasets Design and build prototypes of agentic AI systems, including multi-step workflows that use tools and reason across tasks Apply and experiment with LLM techniques such as retrieval-augmented generation, tool and function calling, fine-tuning, and agent harness design Develop evaluation frameworks to measure the quality, reliability, and safety of LLM and agent outputs Collaborate with mentors, business stakeholders, and engineers to turn business problems into AI solutions and help move them toward production Document methodology, assumptions, and results so work can be reproduced, and present findings and key learnings to technical and business audiences Required qualifications, capabilities, and skills Currently enrolled in a PhD or MS program in Computer Science, Electrical Engineering, Mathematics, Operations Research, Data Science, or a related quantitative field, with an expected graduation date between December 2027 and August 2028 Solid background in at least one of: large language models and agentic AI, natural language processing, speech recognition, or personalization/recommendation, with familiarity with current state-of-the-art methods Strong foundation in statistics, probability, and machine learning, including the ability to design experiments and define metrics that reflect business goals Hands-on exposure to modern LLM application patterns such as prompting, embeddings, retrieval-augmented generation, tool/function calling, structured outputs, and multi-step or agentic workflows Strong Python skills, including deep learning frameworks (PyTorch or TensorFlow), the Hugging Face ecosystem, and core data libraries (NumPy, Pandas, scikit-learn) Clear written and spoken communication for both technical and business audiences, and the ability to work both independently and on a team Preferred qualifications, capabilities, and skills Experience building LLM-powered prototypes that connect models to tools, APIs, documents, databases, or search systems, using frameworks such as LangGraph, Google ADK, LangChain, LlamaIndex, or Semantic Kernel Experience with fine-tuning or post-training (e.g., TRL) and with building evaluation frameworks for LLM or agent outputs Published research in agentic AI, NLP, speech, reinforcement learning, etc at a major conference or journal Ability to write and debug production-quality code Familiarity with the financial services industry #MLCOE_jobs