SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 250,000 - 350,000 / annual
Glean is a Work AI platform that helps enterprises work smarter with AI through intelligent search, AI assistants, and scalable AI agents. The company has evolved from enterprise search into a full-scale AI ecosystem with 100+ SaaS connectors, flexible LLM choice, and robust APIs.
As Tech Lead Manager of AI Evaluations, you will lead a team of applied AI scientists and engineers to build the evaluation strategy and platform that enables Glean to ship reliable and delightful AI experiences. You will own the AI evaluation engineering roadmap end-to-end, from defining success metrics to building the evaluation platform and integrating results into product decisions.
Key responsibilities include:
- Own the evaluation engineering roadmap and technical direction: define success criteria, build the evaluation platform, and integrate results into release and product decisions.
- Build scalable, reusable evaluation infrastructure for datasets, experiments, graders, trace capture and replay, result analysis, and dashboards. Ensure evaluations are reproducible, versioned, cost-aware, and accessible to product and engineering teams.
- Create and maintain representative evaluation sets from expert-curated examples, production signals, synthetic cases, and historical failures. Cover common workflows, edge cases, long-tail requests, adversarial inputs, and important user and domain slices.
- Define a balanced quality measurement system for AI products including correctness, groundedness, retrieval and context relevance, completeness, instruction following, citation quality, user preference, and task success. For agents, evaluate tool selection, multi-step trajectories, policy compliance, recovery, and final task state.
- Develop a fit-for-purpose mix of deterministic checks, model-based graders (including LLM-based autograders), and human evaluation. Establish clear rubrics, validate automated judges against human judgments, and measure evaluator reliability.
- Integrate evaluation into development and operations: fast CI regression checks, pre-release suites, model and prompt comparisons, and sampled production evaluation. Build feedback loops from user feedback and production traces into new test cases.
- Make quality tradeoffs measurable by connecting evaluation outcomes to latency, reliability, and cost. Help teams diagnose failures by slice and trace, prioritize improvements, and make evidence-based decisions.
- Build and lead a high-performing team of applied AI scientists and engineers; coach and develop team members. Partner closely with product, infrastructure, security, and application teams.
Requirements:
- 10+ years of experience in software engineering, ML infrastructure, data systems, or related technical field, with production systems experience.
- 5+ years of engineering or applied AI team management experience, with a track record of technical leadership, developing engineers and scientists, and delivering cross-functional projects.
- Hands-on experience designing, building, or operating evaluation systems for LLM/GenAI products, including model-based evaluators or autograders; practical understanding of how to assess model, retrieval/RAG, assistant, or agent behavior.
- Familiarity with evaluation methods and tradeoffs: curated and production-derived datasets, benchmark creation, deterministic metrics, rubric-based and pairwise judgments, human review, judge validation and calibration, human-model alignment, and statistical analysis.
- Understanding of agent evaluation, including tool calling, multi-turn behavior, trajectory quality, task completion, and recovery from errors; experience with safety testing or red teaming is a plus.
- Strong coding ability in Python or another production language, with sound engineering fundamentals in testing, APIs, data modeling, and reliable distributed systems.
- Experience building scalable data and developer platforms, experiment pipelines, or CI/CD integrations; familiarity with cloud infrastructure, Kubernetes, data stores, and observability is beneficial.
- Strong judgment about measurement quality, including dataset representativeness, judge calibration, reproducibility, uncertainty, and the risk of optimizing to a narrow metric.
- Knowledge of public benchmarks across domains and capabilities is a plus.
- Product-minded and collaborative: able to translate ambiguous quality goals into clear measures, communicate findings to technical and non-technical audiences, and influence roadmaps across teams.
- MS/PhD in Computer Science, AI, or a related field.
