Real-time platform for machine learning

The fast feature engine
developers love to use
Contact Us
Ramp
Melio
Whatnot
Socure
Vital
Moneylion
Apartment List
Found
Pipe
Turo
Feature pipelines in idiomatic Python
Powerful data engineering workflows, without the infrastructure headaches. Powered by Rust.
Built-in scheduling, streaming + caching
Composed & queried in real-time
Toolchain for LLMs
Machine learning infrastructure is painful
Chalk makes it simple for data teams to focus on building the unique products and models that make their businesses thrive.
Chalk has become a critical component of our Risk Intelligence Platform. It expanded Ramp's capabilities with online machine learning and enabled us to scale safely by powering our transaction fraud model and credit underwriting process.
Ryan Delgado Director of Engineering, Data Platform
Deploy to your own infrastructure.
Use your existing database as your online + offline store. No bespoke storage. Everything in your cloud.
High-volume workloads at ultra-low latency.
Chalk’s Compute engine scales horizontally out-of-the-box and executes most complex queries on a Rust-based runtime for maximum performance. 100,000 QPS in < 5ms? We have you covered.
Power real-time decisions with real-time data.
Make better predictions with fresher data. Don’t pay vendors to pre-fetch data you don’t use. Query data just-in-time for online predictions.
user
chalk
query
get_plaid
Apple Watch Series 3
$299.00
Processing...
from chalk.api import ChalkClient ChalkClient().query({ inputs={ Transfer.user.id: 182831, Transfer.amount: 299.00 }, outputs=[Transfer.user.score], })
get_credit_report
name_match_score
is_income_txn
get_user
get_plaid
get_failed_logins
login_counter
200 OK { name: "J.J. Chusterton", age: 28, accounts: [ "Bank of America", "Chase", "First Republic" ] }
Perfect auditability
Know everything you computed and data replay anything.
Parallel Resolvers
This operator executes your Python code in Chalk's massively parallel low-latency runtime environment.
Execution time
4ms
Self time
<1ms
Result size
10MB
Result
500k rows
Groups
4
Groups size
5MB
Runtime
Rust
Resolvers
Features
Output
FeatureValueValue
01
pkey
1
2
02
recent_tx_amts
[130, 24, 87]
[999, 0, 0]
03
fraud_score
0.26
0.13
04
fraud_id
cle2k1
clle09
05
tx_distribution
norm
unif
06
authorization_code
83823
19231
07
authorized
true
true
08
name_match_score
0.99
1.0
Unify training and serving. Iterate faster.
Experiment in Jupyter, then deploy to production.
Prevent train-serve skew and create new data workflows in milliseconds.
Detect, troubleshoot, and eliminate data issues
Track data use, drift, and quality effortlessly with observability—built right in.
Jupyter Notebook
In []
df = client.offline_query( inputs={User.id: user_ids}, outputs=[ User.name, User.credit_report, User.account.bal, ] )
In []
# xgboost train / predict xgb = XGBClassifier( eval_metric="logloss", use_label_encoder=False )
dashboard.chalk.ai
Cache65%
Integrations
Integrate with the tools you already use and deploy to your own infrastructure.
Get Started with Code Examples
Unlock the power of real-time pipelines.
Withdrawal Model
Fraud & Risk
Decide and enforce withdrawal limits with custom hold times.
Income
Credit
Compute income from Plaid transactions.
Bypass the cache with a max-staleness of 0.
Device Data
Predictive Maintenance
Easily listen to streaming data and parse messages with custom logic.
Explore All
Start building with Chalk
Contact Us