Company
Indeed
5 cases from Indeed (Marketplace, General), each digested from a public talk or engineering post and linked to the original. 2 were checked in an online A/B test; 1 report results in production.
Tasks: Evaluate models, Recommend from what the user just did, Learn user and item embeddings, Make metric definitions consistent, Analyze user behavior. Methods: LLM labeling, Human / expert labeling, Uncertainty quantification, Monte Carlo / what-if simulation, Bootstrap, Power analysis / MDE and 7 more. Tools: CatBoost, SHAP and 1 more.
Tech blog: engineering.indeedblog.com
| Calibrating LLM-Based Population Estimates with Human Validation | marketplace | 2026-08 | |
| Bootstrap Confidence Intervals for LLM Evaluation | general | 2026-07 | |
| Distilling Long-Tail User Behavior into Scalable Embeddings for Job Search | marketplace | 2026-06 | |
| Normalized Entropy or Apply Rate? Evaluation Metrics for Online Modeling Experiments | marketplace | 2025-11 | |
| SHAP Plots: The Crystal Ball for UI Test Ideas | marketplace | 2022-04 |