Director: Evaluation and Quality Assurance - Services Special Projects

Summary

We are looking for a Director to build and lead the evaluation and quality assurance function for one of our most ambitious multi-year engineering initiatives. This is a true 0-to-1, first-key-hire role: There is no existing team, tooling, or process to inherit. You will define how we measure quality across an entire product end to end, build the automation and evaluation infrastructure to do it at scale (or borrow other teams' infrastructure), and build and lead the organization that runs it.

Minimum Qualifications

12+ years of experience in data science, machine learning evaluation, quality assurance, or test engineering roles in a technical product environment, including experience building and leading multi-level QA or evaluation organizations through managers. Demonstrated experience building an evaluation or QA function from the ground up — defining metrics, process, tooling, and organizational structure where none previously existed. Extensive experience utilizing LLM-based autograders and rubrics, including rubric design, evaluation set development, validation, calibration, and human-model alignment, and using them to proactively drive product and model improvements. Deep experience in traditional human evaluation methodology, logging, and A/B testing for consumer-facing products at scale. Experience evaluating conversational AI, recommendation, ranking, or search systems, and — ideally — agentic or tool-use systems that take real-world actions, not just generate responses. BS in Computer Science, Engineering, or a related technical field, or equivalent experience. Experience with quality assurance for third-party API integrations. Experience with logging infrastructure and instrumentation for AI product quality measurement. Exceptional written and verbal communication skills, with a track record of presenting quality status, risk, and trade-offs clearly to, and influencing decisions among, senior executive leadership.

Preferred Qualifications

15+ years of experience in data science, machine learning evaluation, quality assurance, or test engineering roles in a technical product environment, including experience building and leading multi-level QA or evaluation organizations through managers.

Description

You will own evaluation and quality assurance for every layer of the product — from whether the system is being correctly triggered for the right set of user requests, through whether it surfaces and ranks the right results, to whether the underlying transactions execute correctly, including integrations with external partner APIs, and financial systems. Quality here isn't a single test suite — it spans conversational understanding, relevance and ranking, and revenue-bearing correctness, all against a large, constantly evolving real-world dataset. You will create and drive the evaluation and quality strategy and implementation for the product. This will extensively utilize automated grading systems and human evaluation programs where appropriate. You will make the calls about where to invest in each. You will define what "correct" and "good" mean for this product, wire that definition into continuous, always-on evaluation pipelines with clear pass/fail quality gates tied to releases, and build the team, tooling, and operating rhythm to catch issues before they reach users. As the first hire in this space, you will operate with significant latitude and ambiguity — setting strategy, process, and technical direction with minimal existing scaffolding, while also building and scaling an organization. You will partner closely with engineering, applied ML, and program management leadership, including senior executive stakeholders, and you will be a key voice in defining what "done" and "shippable" mean for a product still taking shape, as well as iterating on decision points in production after the product ships.

Responsibilities

Design metrics and automated grading systems ("autograders") that scale to large, frequently changing datasets, utilizing those alongside human evaluation, and determining where the line between those should sit. Wiring evaluation results into release processes as explicit quality gates, not just after-the-fact reporting. Catching and preventing quality regressions across an entire product surface — from user-facing correctness to backend transaction integrity. Partnering with engineering and applied ML teams to define what "quality" and "done" mean before a product's own definitions of those things have settled. Navigating ambiguity and building structure — process, tooling, org design — in an environment where none of it exists yet. Making judgment calls under uncertainty about where to invest limited evaluation resources for maximum risk reduction.