AI Harness Engineer
Job Description
Agilent inspires and supports discoveries that advance the quality of life by providing life science, diagnostic, and applied market laboratories worldwide with instruments, services, consumables, application, and both measurement and asset management expertise.
The single point of contact in the pod for architecture, reuse, and quality; builds agents and evals into delivery. This is the engineering edge of the forward-deployed model: building working agentic systems inside real workflows, sitting with the users, learning the domain, and shipping into the tools the experts already live in.
The defining discipline of the role is building agents on the AI platform: consuming certified data products, authoring skills against the registry, integrating through the MCP spine, promoting through eval gates, and ensuring governance.
Responsible for
Architecture and build approach for the pod's use case: agents, RAG, orchestration patterns, and the judgment to know which is appropriate; designs reviewed with the Head of Harness Engineering to stay consistent with reference patterns.
Building on the harness plane as designed: skills authored as minimal, distinct, non-overlapping, independently testable capabilities in the registry; integration through the MCP / A2A spine rather than bespoke plumbing; agent identity and entitlements respected at every data product boundary.
Quality, eval, and observability built into delivery from the first sprint using the shared eval harness: benchmark sets defined with the domain experts, regression gates wired into promotion, and production observability live before launch, not after the first incident.
Risk-tier compliance in practice, including the GxP commit-point pattern on regulated use cases: agent chains operate freely, and any write to a validated record routes through the human-in-the-loop commit point.
Engineering output to the platform: skills, orchestration patterns, and evaluation sets contributed back to the registry as reusable Fabric assets, documented well enough for the second consumer.
Pairing with and upskilling the rotating domain SMEs and Agilent practitioners in the pod, to enable businesses to sustain the engagement and ongoing use of AI solutions, and providing feedback back to the core platform for enhancements.
What success looks like in year one
The pod's agentic system lives in production in the domain workflow, composing registered skills, with documented risk tier, identity propagation, and human-in-the-loop policy.
Eval coverage in place for the use case's critical behaviors, with regression gates passing as a condition of every promotion and measured time-to-detection for behavioral regressions in production.
Multiple assets contributed back to the registry (skills, patterns, or eval sets) and consumed/reused by multiple use cases.
A rotating SME or internal practitioner from the pod is demonstrably more capable than when they arrived, evidenced by their contribution to the build.
Qualifications
Full-stack engineering strength with demonstrated LLM application experience in production:
Experience with building agents, RAG, evaluation, observability, and deployment, not only prototypes.
Working familiarity with MCP or equivalent tool-use protocols, and the architectural taste to build composable systems under a registry discipline rather than one-off integrations.
High agency and tolerance for ambiguity; you are comfortable being the senior engineer in a room of scientists, service leaders, or commercial operators, and you treat their expertise as the specification.
Experience or genuine willingness to operate under regulated-environment constraints; you understand why an audit trail is a feature.
Communication strong enough to demo to a VP and debug with a bench scientist in the same afternoon
Curiosity about AI, its potential and its pitfalls. The field moves monthly, and the people who thrive here are genuinely curious about both sides of it: what these systems can newly do, and where they fail, mislead, or quietly degrade. We want people who read the failure analyses as eagerly as the launch posts, who experiment on their own initiative, and who hold excitement and skepticism at the same time without letting either one win permanently.
Lifelong learner. Whatever expertise a candidate arrives with will be partially obsolete within a year, and that is not a defect of the candidate; it is the condition of the field. We hire people who have reinvented their toolkit before and expect to do it again, who learn in public, and who treat being wrong as information rather than injury. A history of deliberate self-reinvention counts for more than any single credential.
Excellent communication and the ability to influence. Nothing in this organization ships by authority alone. Every role here persuades: domain experts to engage, stewards to share what they know, sponsors to stay honest about value, and functions like Legal, Quality, and Security to move from gatekeeping to partnership. We look for people who write and speak clearly, who adapt their register from bench scientist to Board, and who change minds through credibility and clarity rather than escalation.
Bachelor's or Master's Degree or equivalent.
Typically, at least 8+ years relevant experience for entry to this level.
Additional Details
This job has a full time weekly schedule.Our pay ranges are determined by role, level, and location. Within the range, individual pay is determined by work location and additional factors, including job-related skills, experience, and relevant education or training. During the hiring process, a recruiter can share more about the specific pay range for a preferred location. Pay and benefit information by country are available at: https://careers.agilent.com/locationsAgilent Technologies Inc. is an equal opportunity employer. Qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, protected veteran status, disability or any other protected categories under all applicable laws.