Lead Applied AI Research Scientist

Overview

In this role you will drive applied AI research across the LLM stack within JPMorganChase’s GTAR center, building and evaluating practical AI systems for business use. You collaborate with researchers to publish findings and contribute to IP development, while advancing trustworthy and explainable AI. You will work cross-functionally to integrate AI solutions into business processes and help shape the future of AI at the firm. This is a fast-paced, collaborative environment that values innovation and impact.

Responsibilities

• Advance applied research across the LLM stack (training, adaptation, inference, agentic systems)
• Design, build, and release foundation-model and agentic systems for production-ready business workflows
• Develop evaluation, verification, and monitoring methods for models and agents (faithfulness, hallucination detection, workflow verification)
• Enhance model explainability, reliability, and interpretability
• Provide innovative research solutions to internal project teams
• Document findings and present at conferences; contribute to IP protection
• Collaborate with cross-functional teams to integrate AI into business processes
• Present research outcomes to stakeholders and at industry conferences
• Stay current with AI/ML advancements and foster continuous improvement
• Contribute to the protection of intellectual property

Key requirements

• Ph.D. in computer science, ML, or related fields or Master’s with equivalent applied research and engineering experience
• Ability to build and ship AI/ML systems, ideally involving LLMs
• Proficiency in Python and ML frameworks (PyTorch or JAX)
• Hands-on experience across the LLM lifecycle (fine-tuning, prompting, serving, evaluation)
• Experience in scientific technical writing
• Strong communication and ability to present to non-technical audiences
• Experience in model training/adaptation (supervised fine-tuning, RLHF, distillation, parameter-efficient methods)
• Experience in inference and serving (test-time compute, speculative decoding, quantization, KV-cache)
• Experience in agentic systems (planning, tool calling, retrieval-augmented generation, orchestration)
• Experience in evaluation and faithfulness (benchmarks, LLM-as-judge, hallucination detection, workflow verification)
• Experience in explainability and reliability (attribution, uncertainty, drift monitoring)

• strong communication
• cross-functional collaboration
• ability to present complex results to non-technical audiences
• Python
• PyTorch
• JAX

…