Staff AI Engineer | Agentic Systems

The Role

We're building production-grade agentic systems that audit medical claims end-to-end — reading raw medical records, reasoning over coding and clinical guidelines, and producing defensible findings that hold up to clinical and regulatory review. Reaching human-expert accuracy on noisy, long-context documents is one of the hardest unsolved problems in applied AI, and the field is moving weekly.

We're hiring an L6 AI Engineer to own entire problem areas, not tickets. You'll walk into vague, high-stakes business problems — "our DRG audit findings aren't holding up on appeal," "we need to expand into a new claim type next quarter," "the agent is too slow and too expensive to roll out broadly" — and you'll be accountable for translating them into a technical bet, scoping it with the business, defining the success metric, building the system, and proving it worked. You'll set the technical direction for a problem area and pull other engineers along with you.

What You'll Do

Drive vague business problems to closure. Sit with clinical leads, product, and ops to understand what's actually broken, where the money is, and what "good" looks like. Translate that into a concrete technical problem statement with a measurable target — and push back when the framing is wrong.

Define the metric before you build the system. Decide what you're optimizing (recall on overpayments? appeal-survival rate? cost per case? agreement with senior coders?), how it will be measured, what the baseline is, and what number constitutes shipping. Build the eval harness that produces it. No metric, no project.

Scope and sequence the work. Break an ambiguous initiative into a phased plan with explicit decision points, kill criteria, and dependencies. Decide what's in scope, what's deferred, and what's not worth doing — and communicate that crisply to non-technical stakeholders.

Set the technical direction for a problem area. Choose the agent topology, the context strategy, the model mix, the evaluation regime, the deterministic guardrails. Own the architectural call and the tradeoffs behind it. Other engineers — including senior ones — should be able to build against the foundation you set.

Raise the bar on agent engineering. Lead by example on context engineering, structured outputs, citation grounding, eval discipline, and cost/latency control. Review designs and PRs from other engineers on the team and leave the codebase and the patterns sharper than you found them.

Be the technical interface to the business. Present results to clinical, product, and executive stakeholders. Defend the methodology when findings are challenged. Know the domain well enough to argue with a senior coder about why a code is or isn't supported.

Use AI tooling like a force multiplier. A meaningful fraction of your day will be spent driving Claude Code, Codex, and similar tools to plan, scaffold, refactor, debug, and evaluate. We expect you to be dramatically faster with these tools than most engineers are without them, and to teach the rest of the team to be the same.

What We're Looking For

Required

6+ years of applied ML / AI / software engineering experience with a Bachelor's in CS, Math, Engineering or equivalent — or 4+ years with a Master's / PhD in a similar program. At least two production systems you owned end-to-end from ambiguous problem statement through measured impact, ideally including at least one LLM- or agent-based system.

A track record of driving vague problems to closure. You can point to initiatives where the brief was a paragraph, you scoped it, defined the metric, ran the work, and shipped a result that moved the business — not just a model or a PR.

Strong stakeholder fluency. You can sit with non-technical domain experts (clinicians, coders, ops leads, product), extract what they actually mean, translate it into a technical problem, and translate technical tradeoffs back into terms they can decide on.

Deep, hands-on agent engineering. You've designed agent loops from scratch, decided between single-agent and multi-agent topologies, engineered context (system prompts, tool surfaces, structured outputs, citation grounding), and debugged failure modes that other engineers couldn't.

Eval-first instincts. You don't ship without an eval; you don't believe a number you can't reproduce; and you've built eval harnesses that other engineers on the team now depend on.

Strong Python engineering. Clean abstractions, type discipline, async, tested code — at a level where junior and mid engineers learn from your PRs.

Hands-on experience with at least one major agent SDK — OpenAI Agents SDK, Anthropic SDK / claude-agent-sdk, LangGraph, or equivalent — with strong opinions on the tradeoffs and the scars to back them up.

Fluency with Claude Code / Codex as a power user — able to plan, execute, and debug non-trivial engineering tasks with these tools, including reading their source when needed.

Solid command of VS Code and git — branches, rebases, worktrees, conflict resolution, PR workflows. Not optional.

 

Strongly Preferred

Experience defining and owning a metric that the business actually trusts — precision/recall against expert ground truth, dollar-weighted impact, appeal-survival rate, or equivalent — including the data pipeline behind it.

Prior work on long-context, citation-grounded systems where the model must point to evidence, not just answer.

Healthcare, legal, finance, or any other domain where "mostly right" is unacceptable and where findings get challenged by domain experts.

Experience setting technical direction for a small group of engineers (formal or informal tech lead), including reviewing designs, mentoring on agent patterns, and being accountable for an area's quality.

Familiarity with reasoning models (o-series, Claude extended thinking, Gemini thinking) and a sharp sense of when they earn their cost.

Production experience with caching, observability, and cost control on LLM workloads at scale.

 

Nice To Have

Document understanding (OCR, layout-aware models, table extraction).

Vision-language models, multimodal retrieval.

Experience presenting technical results to executive or external (customer / regulator) audiences.

 

What We Offer

Hybrid role — we have a strong preference for in-office collaboration, with flexibility for exceptional candidates.

Top Medical / Dental / Vision offerings.

FSA / HSA.

Tuition reimbursement.

Competitive salary, 401(k) with company match.

Unlimited PTO.

Meaningful equity.

A flexible, trusting environment where you'll be empowered to do your best work.

Compensation: Base salary for this L6 role ranges $180k–$260k+, based on level assessment, depth of experience, and skill match. Compensation also includes meaningful equity in a fast growing startup and the benefits above.