The opportunity
Waymo is an autonomous driving technology company with the mission to be the world's most trusted driver. Since its start as the Google Self-Driving Car Project in 2009, Waymo has focused on building the Waymo Driver—The World's Most Experienced Driver™—to improve access to…
What you'll do
Own the feature layer between the Growth DB and the models. Work with data: engineering on schema design for ML features, and with data science to turn their feature lists into versioned feature definitions that training and scoring both read from.
Build point-in-time correct training data, so a retrain sees each day as it: was and does not leak future data. Data science has already named this as a gap in today's pipelines.
Take models data science has trained and put them into production: validate against the offline results, wire them into batch scoring, and add input and output checks that stop a bad run before promos go out.
Lead training automation: scheduled retrains, a model registry, and validation before a new model version is promoted.
Monitor models in production: feature drift, score distribution shifts, and model versus random arm results.
Be data science's day-to-day engineering counterpart, so feature and modeling: questions do not queue behind the tech lead.
What they're looking for
- Experience at roughly building production ML systems: feature pipelines, training data generation, batch inference.
- Strong SQL and Python, plus large-scale data processing (Flume, Beam, Spark or similar).
- Designed feature schemas or worked on a feature store, and understands: point-in-time correctness and train/serve skew from having fixed them.
- Productionized and owned the lifecycle of an existing offline model
