Abbott · posted · checked
Principal AI/Agentic Engineer
JavaPythonTypeScript$149k – $299kOn-site · United StatesStaff+
Who can apply
On-site · United States
Location: United States - Illinois - Abbott Park
Abbott is a global healthcare leader that helps people live more fully at all stages of life. Our portfolio of life-changing technologies spans the spectrum of healthcare, with leading businesses and products in diagnostics, medical devices, nutritionals and branded generic medicines. Our 122,000 colleagues serve people in more than 160 countries.
JOB DESCRIPTION:
About Abbott
Abbott is a global healthcare leader that helps people live more fully at all stages of life. Our portfolio spans diagnostics, medical devices, nutritionals, and branded generic medicines, and our colleagues serve people in more than 160 countries.
The Opportunity
A hands-on software engineering leadership role accountable for building production-grade agentic systems on Abbott's enterprise AI Platform. This role turns business and product requirements into reliable single- and multi-agent solutions, reusable engineering patterns, and measurable outcomes. The role is Director-mapped and begins as a player-coach: it leads delivery and technical standards while building and managing a team of agentic engineers.
What You'll Do
Design, implement, and operate production agentic systems, including supervisor, planner/executor, and multi-agent patterns with explicit boundaries, failure handling, and human oversight.
Build durable, resumable workflows with persistent state, checkpoints, replay, idempotency, compensation, and human-in-the-loop controls (e.g., LangGraph- or Temporal-style durable execution).
Engineer tool and function integrations with Abbott enterprise systems, including MCP-style tool contracts, schemas, authorization, retries, timeouts, compensation, and failure containment.
Develop context-engineering strategies covering retrieval, chunking and indexing, memory hierarchies, context-window budgeting, caching, structured outputs, and grounding.
Create evaluation systems with curated datasets, task-success and safety metrics, LLM-as-judge and programmatic evaluators, regression gates in CI, and online, A/B, and shadow evaluation.
Instrument agent applications through the platform's model and agent registries and tracing services so prompts, tool calls, intermediate steps, quality, cost, and latency are observable and auditable.