Build production AI agents at JPMorganChase within the Corporate & Investment Bank Technology organization. You will work on the NEO agent runtime platform on AWS and Azure, designing, productionizing, and operating LLM-powered agents that support use cases across the Corporate & Investment Bank and Payments.
NEO already runs a federated portfolio of production agents, including forecasting, anomaly detection, and log analysis, with additional capabilities such as sales fulfillment and voice-of-client close behind. In this role, you will extend the platform by shipping new production agents and strengthening the underlying retrieval, memory, evaluation, and governance patterns.
What you’ll do
- Design and ship production agents on NEO across the federated portfolio, owning work from prototype through production.
- Build production-ready retrieval including chunking, ranking, and grounding strategies to keep answers accurate and auditable.
- Design agent memory with episodic and semantic memory organized as memory nodes, including recall, summarization, and decay policies tuned per use case.
- Own organizational context management so each agent reasons over only permitted information using entitlement-, lineage-, and tenant-aware context.
- Compose multi-agent workflows using A2A, and integrate tools and data through MCP servers (Bitbucket, Confluence, Databricks, Kubernetes, Snowflake, Splunk).
- Build and run evaluations including task-level and end-to-end agent evaluations, regression suites, LLM-as-judge, and quality/safety gating before release.
- Deploy and operate solutions on AWS and/or Azure using strong SDLC, security, resiliency, and observability practices.
- Partner with product and business teams across the Corporate & Investment Bank and Payments to turn use cases into shipped, supported agents.
What you bring
- MS in Computer Science, Statistics, Mathematics, Machine Learning, or related field (or equivalent experience).
- 7+ years of development experience, with at least 4 years working on AI/ML solutions.
- Hands-on experience building LLM-powered or agentic applications in production, including tracing, evaluations, and guardrails.
- Strong programming skills in Python, with deep knowledge of data structures, algorithms, machine learning, data mining, information retrieval, and statistics.
- Practical RAG experience, including retrieval quality, embeddings, and vector stores.
- Expert knowledge of at least one: AWS, Azure, Kubernetes.
- Knowledge of data management and data model design; real-time processing using SQL (e.g., Postgres) and NoSQL stores (e.g., OpenSearch, Redis).
- Excellent communication skills with the ability to collaborate with senior technical and business stakeholders.
Preferred capabilities
- Graph RAG experience (combining knowledge-graph traversal with vector search).
- Experience with agent frameworks or runtimes, including A2A or MCP.
- Experience with agent memory design (memory nodes, episodic/semantic memory) and organizational context management.
- Knowledge of knowledge graphs and graph databases used for retrieval.
- Understanding of LLM fine-tuning and small language model inference.
- Ability to build full-stack products using modern JavaScript/TypeScript frameworks (e.g., Next.js, Svelte) for agent UIs (AG-UI / NEO UI SDK).
- Experience in financial or payments domain at a large institution (Investment Banking, Markets, Securities Services, or adjacent).
- Knowledge of high-performance languages such as Go or Rust.
Tools and technologies
NEO, AWS, Azure, Python, Databricks, GenAI Gateway, MCP, Bitbucket, Confluence, Kubernetes, Snowflake, Splunk, SQL, Postgres, NoSQL, OpenSearch, Redis, A2A, LLM-as-judge, vector stores, embeddings, Graph RAG, knowledge graphs, graph databases, JavaScript, TypeScript, Next.js, Svelte, AG-UI, NEO UI SDK, Go, Rust.
Role details
Location: Jersey City, NJ (onsite)
Compensation: USD 171,000 - 260,000 per year