Aqueduct is an open-source production data science platform that abstracts away the software engineering required for high-quality, consumable data science.
Get StartedGitHub
Aqueduct automates away the engineering required to deploy, connect, and monitor data science projects in production — whether it's a simple model running locally or a large-scale prediction task in the cloud.
Rather than spending days or weeks, you can deploy a model with two lines of Python -- under the hood, Aqueduct abstracts away all of the low-level cloud infrastructure needed to make it real.
Aqueduct comes with a pre-built suite of connectors to data warehouses, databases, and business systems, allowing you to publish predictions wherever they're needed.
Prediction pipelines can break — often silently — for lots of reasons, including schema changes, data distribution shifts, and model drift. With Aqueduct's metrics & checks, you get detailed visibility into the ongoing model performance and immediately get notified if things go awry.
Data scientists should have full control over their prediction pipelines, from definition to deployment to monitoring — but today, they don't. Aqueduct enables data scientists to be self-sufficient and take control of their predictions by eliminating the engineering required to deploy & manage ML models.
Today, data engineers need to manage complex & tedious deployment and debugging processes in order to enable data scientists to succeed. Aqueduct removes the need for data engineers to manage custom deployment processes and retrieve context around when and how ML models break.
MLOps slows down every data team by demanding data engineers and data scientists both be involved with every minor change to your prediction pipelines. Aqueduct eliminates this operational inefficiency by streamlining prediction delivery for everyone on your team.