Train models with  production features

Generate point-in-time correct training datasets from the same feature definitions you use in production and serve production features from your training data. Keep training and inference aligned and eliminate skew.

TALK TO AN ENGINEER
hero gradient image

Trusted by the teams building next generation of AI + ML

logologologologologo

Training-serving consistency

Evaluate training with the same feature definitions and resolvers used online to eliminate skew and rewrites.

High-throughput, large scale

Generate billions of rows of training datasets through distributed offline queries without provisioning infrastructure.

Point-in-time correctness

Evaluate features using only the data available at each historical moment to avoid leakage and ensure accurate training inputs.

Branch-based experimentation

Create feature branches, generate comparison datasets, and seamlessly promote winners to production.

Deterministic & reproducible

Produce versioned datasets for reproducible results.

Chalk is a key part of our underwriting pipeline, letting us test new code against production pipelines without disruption. We can quickly iterate on new features and enhancements, helping us offer more flexible capital products than anyone else.

Nate Wiger CTO

company logo

Generate point-in-time datasets from production features

product section resource

Point-in-time evaluation ensures features are computed using only the values available at each historical moment. This removes leakage, produces reproducible datasets, and allows offline experiments to reflect production behavior.

Explore offline queries

Integrate with your model training workflow

product section resource

Training datasets generated with Chalk integrate directly into your model training workflow. Chalk preserves the schemas and evaluation semantics used in production so models train on the same feature logic they use at inference.

Model training

Backfilling with chalk.now

from chalk import Now, online

@online
def get_age(birthdate: User.birthdate, now: Now) -> User.age:
    return (now.date() - birthdate).days // 365

Evaluate features as if it were any point in time. Train on historically accurate windows, then serve the same features in production with no drift or leakage.

Use chalk.now

The latest at Chalk

Train your models with production features

Talk to an engineer and see how Chalk can power your production AI and ML systems.

Talk to an engineer

Explore more of Chalk’s data platform