Staff AI Engineer | Agentic Systems

Machinify

14h•$180,000 - $260,000•Hybrid

About The Position

Machinify is hiring an L6 Staff AI Engineer to own problem areas for production-grade agentic systems that audit medical claims end to end. These systems read raw medical records, reason over coding and clinical guidelines, and produce findings that can withstand clinical and regulatory review. The role addresses difficult applied AI challenges involving noisy, long-context documents and human-expert-level accuracy. The engineer will take ambiguous, high-stakes business problems from discovery through measurable results: work with clinical, product, and operations stakeholders to understand needs, define technical targets and evaluation methods, scope and sequence initiatives, build systems, and verify their impact. The role sets technical direction for a problem area, makes architectural and tradeoff decisions, establishes engineering patterns, and helps other engineers build effectively against the resulting foundation. The position also serves as a technical interface to business stakeholders, including presenting and defending results and engaging substantively with clinical coding questions. The engineer is expected to use AI-assisted development tools such as Claude Code and Codex extensively and help raise the team's effectiveness with them.

Requirements

At least 6 years of applied ML, AI, or software engineering experience with a bachelor's degree in computer science, mathematics, engineering, or an equivalent background; alternatively, at least 4 years of experience with a master's or Ph.D. in a similar program.
Ownership of at least two production systems end to end, from an ambiguous problem statement through measured impact; ideally, at least one system is LLM- or agent-based.
Demonstrated ability to scope vague initiatives, define metrics, execute the work, and ship results that move the business.
Ability to work with non-technical domain experts, elicit their needs, translate them into technical problems, and explain technical tradeoffs in decision-ready terms.
Hands-on experience designing agent loops, choosing between single-agent and multi-agent topologies, engineering context and tools, using structured outputs and citation grounding, and debugging agent failure modes.
Evaluation-first approach and experience building evaluation harnesses that other engineers rely on.
Strong Python engineering skills, including clean abstractions, type discipline, asynchronous programming, and tested code.
Hands-on experience with at least one major agent SDK, such as OpenAI Agents SDK, Anthropic SDK or claude-agent-sdk, LangGraph, or an equivalent.
Power-user fluency with Claude Code or Codex, including planning, executing, and debugging non-trivial tasks and reading tool source when needed.
Solid command of VS Code and Git, including branches, rebases, worktrees, conflict resolution, and pull request workflows.

Nice To Haves

Experience defining and owning a business-trusted metric, such as precision or recall against expert ground truth, dollar-weighted impact, or appeal-survival rate, including the supporting data pipeline.
Experience with long-context, citation-grounded systems in which models must point to evidence.
Experience in healthcare, legal, finance, or another domain where findings are challenged by domain experts and mostly-right results are insufficient.
Experience setting technical direction for a small group of engineers, formally or informally, including design reviews, mentoring on agent patterns, and accountability for quality.
Familiarity with reasoning models such as o-series, Claude extended thinking, and Gemini thinking, and the ability to assess when their cost is worthwhile.
Production experience with caching, observability, and cost control for LLM workloads at scale.
Experience with document understanding, including OCR, layout-aware models, or table extraction.
Experience with vision-language models or multimodal retrieval.
Experience presenting technical results to executive or external audiences, such as customers or regulators.

Responsibilities

Work with clinical, product, and operations stakeholders to identify business problems, clarify what is broken, and translate needs into measurable technical problem statements.
Define success metrics, baselines, measurement methods, and shipping thresholds before building systems.
Build evaluation harnesses and ensure results are reproducible.
Break ambiguous initiatives into phased plans with decision points, kill criteria, and dependencies; communicate scope and sequencing to non-technical stakeholders.
Set technical direction for agent systems, including agent topology, context strategy, model mix, evaluation regime, and deterministic guardrails.
Lead by example in context engineering, structured outputs, citation grounding, evaluation discipline, and cost and latency control.
Review other engineers' designs and pull requests, and improve the team's code and engineering patterns.
Present results to clinical, product, and executive stakeholders; defend methodologies when findings are challenged.
Apply domain knowledge to reason about whether medical codes are supported by evidence.
Use Claude Code, Codex, and similar AI tools to plan, scaffold, refactor, debug, and evaluate engineering work, and teach teammates to use them effectively.