Transform. Serve. Scale.
Feature Platform for Real-Time Machine Learning
Power all your models with production-ready batch, streaming, and real-time data pipelines
From the creators of Uber Michelangelo, the ML Platform powering every model at Uber. Tecton is trusted by leading ML teams.
More than just a feature store
The All-in-one System to Build, Automate, and Centralize Feature Workflows for Production ML
Production Pipelines in Minutes
Create robust data pipelines from just a few lines of code — Tecton automatically builds, orchestrates, and maintains them so your teams don’t have to
Increased Model Accuracy
Automate robust pipelines that materialize fresh features consistently across training and serving environments to ensure improved model performance
Improved Collaboration
Standardize machine learning data workflows in one platform so teams can easily share and reuse features across use cases
Mission-Critical Reliability
Serve features in production at extreme scale for batch, streaming, or real-time ML with the confidence that systems will always be up and running
The problem we solve
The problem we solve
Data Engineering for Real-Time ML Is Hard
- Training / serving skew
- Point-in-time correctness
- Productionizing notebooks
- Real-time transformations
- Melding batch + real-time data
- Complicated backfills
- Latency constraints
- Ensuring service levels at scale
A complete platform
Design and Manage the Entire ML Feature Lifecycle
From simple declarative transformation logic to fresh feature values you can store, serve, and monitor, in real-time.
Explore the product
Feature Management: Discover, Use, Monitor, and Govern End-to-end Feature Pipelines
With Tecton, anyone on the team can discover and use existing features all the while monitoring associated data pipelines, serving latencies, processing costs, and underlying systems of their machine learning applications.
Feature Logic: Design and Define ML Features
By using the SDK in a notebook or any other Python environment, users can define feature logic in SQL, PySpark, SnowPark, or Python, and rely on Tecton to execute complex data transformations like time-window aggregations or generate training data with accurate backfills.
# Specify inputs, entities and compute configuration
@batch_feature_view(
description="Whether the user has a good credit score, updated daily",
sources=[credit_scores],
entities=[user],
feature_start_time=datetime(2020, 10, 1),
batch_schedule=timedelta(days=1),
)
# Define transformation logic
def user_credit_quality(credit_scores):
return f'''
SELECT
USER_ID,
IF (CREDIT_SCORE > 700, 1, 0) AS USER_HAS_GOOD_CREDIT,
TIMESTAMP
FROM
{credit_scores}
'''
# Specify inputs, entities, and compute configuration
@stream_feature_view(
description="Mean transaction amount of last hour, 24h and 72h, updated every 10 min",
source=transactions,
entities=[user],
feature_start_time=datetime(2020, 10, 1),
# Define logic for time-window aggregations
aggregation_interval=timedelta(minutes=10),
aggregations=[
Aggregation(column='AMOUNT', function='mean', time_window=timedelta(hours=1)),
Aggregation(column='AMOUNT', function='mean', time_window=timedelta(hours=24)),
Aggregation(column='AMOUNT', function='mean', time_window=timedelta(hours=72)),
],
)
# Define query to use in the transformation
def mean_transaction_amount(transactions):
return f'''
SELECT
USER_ID,
AMOUNT,
TIMESTAMP
FROM
{transactions}
'''
# Indicate schema of incoming request-time data (e.g. live transaction data)
transaction_request = RequestSource([Field('amount', Float64)])
# Specify inputs and schema of the output
@on_demand_feature_view(
description="Whether the current transaction amount is higher than the user's weekly average.",
sources=[transaction_request, historical_metrics],
schema=[Field('transaction_amount_is_higher_than_average', Bool)],
)
# Define transformation logic
def transaction_amount_is_higher_than_average(transaction_request, historical_metrics):
result = {}
result['transaction_amount_is_higher_than_average'] = (
transaction_request['amount'] >= historical_metrics['amount_7d_mean'])
return result
Feature Repository: Register and Collaborate on Feature Definitions
Tecton’s feature repository lets users manage feature definitions as files in a git-like repository. With Tecton, users define features in code, version control them in git, unit test them, and roll them out safely using Continuous Delivery pipelines. With Tecton, bring battle-tested DevOps software practices to feature engineering.
Feature Engine: Transform Raw Data Into Fresh Feature Values
Tecton integrates with existing data processing and storage infrastructures to automatically compile the underlying data pipelines that compute batch, streaming, or real-time features, insulating the end user from their complexity.
Feature Store: Store and Serve Fresh Feature Values
Based on a user’s pre-defined requirements, Tecton helps organizations scale compute, storage, and serving independently to adjust to usage patterns, leverages an offline store for large-scale and low-cost retrieval (training) and an online store for low-latency retrieval (online serving). Tecton provides uninterrupted access to fresh features on demand.
Customer Stories
Battle tested for the enterprise
Reliability and Scale You Expect From a Production System
Fully Managed and Cloud Native
Tecton makes it easy to deploy and operate machine learning with a managed, cloud native service.
Built for Availability and Scale
Tecton is built for scale, delivering median latencies of ~5ms and supporting over 100,000 of requests per second.
Integrates with Your Existing Stack
Tecton is not a database or a processing engine. It plugs into and orchestrates on top of your existing storage and processing infrastructure.
Secure and Compliant
Tecton authenticates users via SSO and includes support for access control lists. We support GDPR compliance in your ML applications, and are SOC2 Type 2 certified.
