Continuously improve your ML-powered products
Gantry helps you improve your ML products with analytics, alerting, and human feedback
Loved by ML engineers
At Flowrite, we build AI systems that utilize some of the latest innovations in NLG. With Gantry’s help, we’ve been able to falsify important hypotheses in our experiments in record time and that has allowed us to direct our product development with speed and confidence.
Mo’ models, mo’ problems. If monitoring one model is hard, monitoring thousands is much harder. Gantry helps us tackle the complexity of understanding how models are performing in production in a B2B AI company, and what that means for our customers.
Gantry gets me fast, fine-grained insights into my models’ behavior in the wild. I am able to skip notebook based data exploration, as I can poke around and explore the big patterns and odd corners right in Gantry’s UI, and then pull down the data for a deeper dive or more modeling.
A simple, powerful SDK for model instrumentation
Gantry's SDK gives you easy access to all of your production data and metrics, just by adding a few lines of code.
prediction = model.predict(inputs)
gantry.log_record(
application="Model Service",
inputs=inputs,
outputs=prediction,
environment="production",
feedback_id=my_uuid
)labels = fetch_from_labeling_service(inputs)
gantry.log_records(
application="Model Service",
inputs=inputs,
feedback=labels
)
user_accepted = on_user_click()
gantry.log_record(
application="Model Service",
feedback_id=my_uuid,
feedback=user_accepted,
)
Analytics, exploration, and visualization to improve your models
Use our dashboard to discover user cohorts, data slices, and edge cases where your model is biased or underperforms in production. In Gantry, every dashboard query is also available in our Pandas-like SDK, so you can continue to explore all of your production data in a Jupyter notebook.
Make sense of unstructured data
Uncover a wide range of patterns in your text, images, or other messy data without writing code using our off-the-shelf projections. Extend Gantry with your own custom projections to build a monitoring and data analysis suite for your unique model.
Detect concerning model behavior before it becomes a problem
Simple model monitoring and alerts. Compute metrics about data quality, data drift, model performance, and user satisfaction. Set thresholds, receive slack notifications about degradations, and even get daily updates without configuring anything.
Deploy securely at any scale
Deploy securely
- Gantry is built with enterprise-grade authentication. We’re SOC 2 compliant.
Handle any data volume
- Gantry is built on cloud-native technologies and designed to scale to terabytes of data.