Real-Time Serving
Compute features at inference time,
directly from the source
On-demand features are computed real-time by Chalk resolvers, expressed as ordinary Python or SQL. Chalk builds a query plan across your sources and returns computed values in milliseconds. The same definitions power training and production, eliminating train/serve skew.
Explore Chalk’s real-time serving
Composable Python or SQL
A resolver is an ordinary function that can call other resolvers, and Chalk works out the execution. No DSL, no proprietary pipelines.
Eliminate train-serve skew
One feature definition serves training and production, so both see the same values.
Freshness you set
Declare how stale a value may be. Chalk serves from cache inside that window and recomputes outside it.
Computed on demand
Run on a cron, off a stream, only when queried, or as a materialized aggregation.
Built for real-time inference
Cached and materialized values return in single-digit milliseconds.
Traced & testable
Every resolver run is traced end-to-end like ordinary Python. Streaming resolvers add a test harness.
Our entire Search Ranking feature stack is now on Chalk, computing and retrieving features for inference in just milliseconds. We’re transitioning all of our other models to Chalk, including streaming features for real-time, personalized insights.
One definition,
every environment
A resolver reads from the systems you already run, computes a value, and returns it in the same shape whether the caller is a training job or a production request.
Resolver Overview
Keep up with Chalk
What we've been up to and where to find us next.
Most Innovative Companies
See why Fast Company named us on their list.
Register for the upcoming webinar
Learn how to give agents context for evals and production
Raising the Talent Bar
How Chalk filled 40+ roles in one quarter without lowering the bar
Introducing Chalk Notebooks
Read more about our latest product announcement
Define once and serve everywhere
Talk to an engineer about serving better context at inference.