Senior Associate- Applied AI/ML
JPMorgan Chase · Jersey City, NJ, United States · Full time
Posted 2026-11-17
What this role cares about
- Model Development
- Natural Language Processing (NLP)
- Data Processing
- Feature Engineering
- Model Improvement
- Cloud Operations
- Data Pipeline
- Model Deployment and Integration
- Model Monitoring and Maintainance
- Machine Learning Operations (MLOps)
- Scalability
- Product Development
Tools: TensorFlow, PyTorch, AWS, GitHub, Jira, Python, Reinforcement Learning, APIs, Claude, LLM, NLP
Job description
Job Responsibilities Develop advanced agentic AI solutions across NLP, speech analytics, time series, reinforcement learning, and recommendation systems. Design robust agent architectures combining LLM reasoning with tools, structured data, and APIs—spanning state, memory, and context management, plus loop engineering (plan/act/observe, verification, termination, and fallback/escalation). Engineer reliable agent-driven workflows emphasizing correctness, traceability, and control-aware behavior (guardrails, approvals, auditable decision paths). Build knowledge-centric reasoning layers, including knowledge graphs and hybrid retrieval (RAG + graph + structured sources) to improve grounding and accuracy. Drive specification-driven development: author specs and contracts (schemas, validators, tool/skill interfaces) and build evaluation/regression harnesses. Advance agent quality via recursive self-improvement through automated evaluation and critique loops, red-team feedback, skill/prompt instruction optimization, and outcome-driven dataset curation (human-in-the-loop as needed). Design and implement multi-agent orchestration pipelines for SRE/observability use cases — integrating Dynatrace APM (problem APIs, golden signals, entity resolution), Splunk/SPL (Atlas CloudWatch and OnPrem query patterns), infrastructure metadata , code search, and JIRA ticket automation. Author agent specs as formal contracts: define phase-by-phase protocols, STRICT CONSTRAINTS, backflow/escalation signals, auth-error handling, and per-phase checkpoint schemas. Implement causal reasoning frameworks for evidence-based RCA and confidence scoring. Build A/B parity evaluation frameworks and regression harnesses to measure agent behavior changes across releases; define merge-gate criteria and scoreboard metrics. Develop MCP (Model Context Protocol) server integrations and plugin architectures for AI coding assistants (Claude Code). Implement per-application customization knowledge banks and agent memory systems that persist context across investigation phases. Coach and mentor AIML team members, setting a high bar for engineering rigor and research depth. Required Qualifications, Capabilities, and Skills Studies in a quantitative discipline ( CS/EE/Math/OR/Optimization/Data Science) or equivalent industry/research experience Demonstrated expertise building agentic AI systems, including several of: memory/state/context management, tool orchestration and workflow reliability patterns, loop engineering, spec-driven development, and prompt/skill instruction optimization. Strong hands-on experience with ML/DL methods and toolkits (e.g., PyTorch/TensorFlow plus core Python data/ML stack). Ability to design experiments and evaluation frameworks with metrics aligned to business outcomes (quality, reliability, latency, cost, safety). Experience with scalable data and model workflows (training and/or inference) and strong software engineering practices. Strong communication skills to explain technical concepts to both technical and business audiences. Hands-on experience with at least one observability/APM platform (Dynatrace, Splunk, Datadog, New Relic, or equivalent) — including writing queries, interpreting golden signals, and using APIs for automated incident triage. Proficiency in Python for pipeline scripting, automation, and agent tooling; comfort with JSON schema design and structured output validation. Familiarity with AWS services relevant to observability and agentic workloads (ECS, EKS, Lambda, CloudWatch, ALB, Step Functions) Familiarity with MCP (Model Context Protocol) server development or plugin authoring for AI coding assistants (Claude Code, GitHub Copilot, Cursor, or equivalent). Preferred Qualifications, Capabilities, and Skills Knowledge in search/ranking, reinforcement learning, or meta-learning (especially for agent routing, policies, and self-improvement). Experience with knowledge graphs, entity resolution, and ontology design. Experience with A/B experimentation and metric-driven product development; CI pipelines and unit/integration testing. Hands-on experience with Splunk SPL — writing complex search queries, index resolution, field extraction, and alert patterns for production observability. Experience with Dynatrace APIs — entity queries, problem feeds, golden signal retrieval, and synthetic monitor management. Experience with SRE practices: incident response workflows, root cause analysis methodology, SLOs/SLIs, on-call operations, or production readiness reviews Experience with GraphRAG, multi-hop retrieval, or knowledge-graph-augmented agent reasoning. Forward Deployed Engineering or Spec-Driven Development experience — translating operational workflows into formal agent specs with schemas, validators, and compliance audit trails. Experience building evaluation harnesses for long-horizon agentic tasks: golden-dataset construction, LLM-as-judge scoring, regression detection, and parity testing across agent versions.