Staff Machine Learning Engineer, Agent Eval Platform

Job Description

The Role

Moveworks' AI agents don't just generate text — they act. They plan, call tools, and change real state in enterprise systems on behalf of 5.5 million employees. Building agents is no longer the bottleneck. Improving them is: knowing whether a change made the agent better, why a conversation failed, and which part of the system should be fixed.

You'll build the loop that makes our agents improve themselves — finding failures in real use, deciding which part of the agent harness is responsible, checking every proposed fix on cases the fixer never saw, and turning each fixed failure into a permanent test. We improve the system around the model: instructions, tool descriptions, routing, knowledge and memory. We don't train the model itself.

The work is blend of ML and systems engineering. You'll decide whether a signal is trustworthy enough to act on, and build the infrastructure that produces that signal reproducibly across hundreds of customer organizations.

This isn't a model-training role, and it isn't a testing role. It's applied ML at a point where the methodology genuinely isn't settled: LLMs judging LLMs is an open research problem, and we're working it against agents that take real, irreversible actions in stateful, multi-tenant enterprise environments.

What you get to do in this role:

Judge design and calibration

• A shared base judge with per-item rubrics expressed as configuration next to the dataset — so eval authors express intent, rather than forking a prompt per eval
• Splitting the problem correctly: deterministic validators for checkable world state ("was the ticket created, with the right item, routed to the right approver?"), and an LLM judge for the parts that are genuinely fuzzy — was the clarifying question appropriate, was policy followed, was the path efficient
• Scoring that reports its own confidence , so uncertain judgements route to a human instead of quietly becoming trusted signal
• A standing calibration loop against human-labeled trajectories, run in partnership with our annotation team — they own the human labeling, you own the calibrated judge artifact. How consistently humans agree with each other sets the ceiling on how good any judge can be, so raising that ceiling is part of the job
• Guarding against correlated blind spots: our user simulator and our judge are both LLMs, and they can be wrong in the same direction
• Offline↔online divergence: when simulation and production disagree, being the person who can say why, and keeping the suite re-seeded from new production failures so it can't quietly overfit

Self-improving agent harness

This is where the pillar is headed, and a large part of why the seat exists. You'll build the components that let the agent harness get better from its own use, without changing the model underneath.

• Finding what to improve: surfacing where the agent falls short in real use, including failures that look like success
• Synthetic data and environments: generating scenarios and the seeded enterprise worlds they run in — with expected outcomes and checks — from real failures or ahead of what customers ask for, and making sure what gets generated is realistic, hard enough, and doesn't leak the answer
• Deciding where a fix belongs: understanding why the agent failed and which part of the harness should change: its instructions, tools, routing, knowledge or memory
• Proposing improvements: optimizing those components against simulation rather than against production traffic
• Checking before anything ships: verifying every proposed change on cases it wasn't tuned on, catching regressions an average would hide, and keeping the part that proposes a change separate from the part that judges it
• Evals that grow: every fixed failure becomes a permanent test, so the suite gets stronger the more the agent is used
• Safe across many customers: learnings stay scoped to one organization where they must, and admins approve anything org-specific
• The infrastructure that runs it: a loop that runs reproducibly, at scale, across hundreds of customer organizations
• The guardrail that keeps it honest: end-state outcomes are what we hold the agent to

Qualifications

Required:

• 5+ years in applied ML or ML-adjacent engineering, with systems you shipped, operated, and others relied on
• Experience turning subjective human judgement into a measurement that holds up — one that a team gated decisions on, validated against human labels. This is the core of the job
• You've improved a multi-component system through measurement — pipelines, retrieval, prompts, policies, or routing — not only by changing model weights
• Strong applied ML fundamentals, and comfort treating LLMs as a component you evaluate and prompt rather than one you train
• Strong Python and production engineering: services, data pipelines, and designs that make results reproducible (versioned data, deterministic replay, recorded inputs)
• Ability to think and communicate clearly about complex problems — a large part of this job is convincing engineers that a number means what you say it means, and being right
• A high degree of ownership and a bias toward shipping at startup pace
• Comfort with ambiguity, and the judgement to know when a measurement is good enough to act on

Experience in at least 3 of these:

• LLM-as-judge or automated evaluation design, and calibrating it against human judgement
• Human annotation programs: rubric authoring, label quality, and annotator throughput as a real constraint
• Experimentation rigor: held-out sets, decision rules set up front, variance and noise, regression detection
• Failure analysis at scale: classifying, clustering, and attributing errors in logs or traces
• Replay, simulation, or sandboxed test infrastructure
• Synthetic data generation for evaluation: generated scenarios, seeded environments, and quality controls on what gets generated
• Prompt and harness optimization as an engineering discipline — versioned, tested, and measured, not tuned by vibes
• Search ranking, recsys, or online experimentation evaluation — golden-set staleness, offline/online divergence, side-by-side rater agreement
• High-volume trace or event data pipelines

At the Staff level, additionally: you'd own the measurement architecture of the self-improving loop end to end — judging, deciding where fixes belong, and checking them — set the measurement standard the organization uses to decide whether a change ships, be the technical counterpart to the team that builds the agent, and mentor engineers across the team on evaluation methodology.

Bonus:

• Evaluated LLM agents: multi-step trajectories, tool calls, actions in stateful environments
• Generated complete agent environments — tasks, stateful worlds, and verifiers — automatically, at scale
• Published or shipped evaluation methodology others adopted
• Built eval tooling used by teams outside your own, or an eval product customers use
• Worked on agentic systems that take actions, not just generate text

For positions in this location, we offer a base pay of $176,100 - $308,200, plus equity (when applicable), variable/incentive compensation and benefits. Sales positions generally offer a competitive On Target Earnings (OTE) incentive compensation structure. Please note that the base pay shown is a guideline, and individual total compensation will vary based on factors such as qualifications, skill level, competencies and work location. We also offer health plans, including flexible spending accounts, a 401(k) Plan with company match, ESPP, matching donations, a flexible time away plan and family leave programs.

Compensation is based on the geographic location in which the role is located and is subject to change based on work location.

Additional Information

Work Personas

…