AI Engineer - Data and Evals

Glean•Mountain View, CA

3h•$200,000 - $280,000•Hybrid

About The Position

Glean is hiring an AI Engineer to build data and evaluation science systems for its AI products. The team develops evalsets, agentic LLM judges, benchmarks, evaluation infrastructure, and enterprise simulation environments to measure product quality and guide improvements across the AI stack. The work draws on real production usage and supports improvements to agent harnesses, tool use, skills, model routing, and post-training of task-specific models. The role partners with R&D, product, and go-to-market teams to bring research into production and translate evaluation findings into better products and customer outcomes. This is a hybrid role requiring four days per week in the Mountain View or San Francisco office.

Requirements

At least 5 years of software engineering experience.
Strong Go and/or Python skills and backend fundamentals, including APIs, data modeling, testing, and distributed pipelines.
Experience building datasets and benchmarks for LLM use cases, RAG, or agents, including evalset curation and synthetic data generation.
Familiarity with evaluation methodologies and their tradeoffs, including deterministic metrics, LLM-as-a-judge, pointwise versus pairwise grading, human labeling, and judge calibration.
Experience building evaluation or experiment infrastructure that runs reliably at scale, including CI/CD integrations and observability tooling such as OpenTelemetry.
Strong analytical skills, including investigating traces and failure cases, writing SQL, and using data and metrics to explain changes in quality.

Nice To Haves

TypeScript experience.
Experience with simulation environments and/or model post-training techniques.

Responsibilities

Build evalsets and benchmarks from expert-curated examples, production data, synthetic data, and past failures, covering edge cases and key user slices.
Build and scale the evaluation platform and agent observability, including experiment pipelines, trace capture and replay, and dashboards, with reproducible and affordable runs.
Build agentic LLM verifiers for correctness, groundedness, citation quality, and task success, and calibrate them against human labels using clear rubrics.
Run evaluations at every stage, from CI and pre-release suites to sampled production traffic, and report results with confidence intervals.
Build enterprise simulation environments with simulated users and realistic company data to test multi-step agent tasks.
Diagnose failures by slice and trace, and turn evaluation results into concrete fixes while balancing quality, latency, and cost.
Partner with R&D, product, and go-to-market teams to bring research into production and translate evaluation findings into product and customer improvements.