Ketan Umare, Sandra Youssef
September 27 (11:15 AM - 12:05 PM PT)
September 27 (4:50 PM - 5:40 PM PT)
September 28 (5:00 PM - 5:50 PM PT)
Ketan Umare, Katrina Rogan
September 29 (2:45 PM - 3:35 PM PT)
Battle-Tested and Truly Open-Source
Resiliency is what we’re after
Flyte has been battle-tested at Lyft, Spotify, Freenome, and others. It is entirely open-source with an Apache 2.0 license under the Linux Foundation with a cross-industry overseeing committee.
Write Code, Not YAML!
Leave the infrastructure to us
Configuring Machine Learning and Data workflows can get complex and error-prone with YAML. Flyte provides intuitive, user-friendly SDKs to get the workflows up and running in no time.
Start Small, Scale Big
Deploy to a Kubernetes cluster
Run on demand
Flyte is a Kubernetes-native workflow automation platform to unify data and ML processes. The containerized, microservices-based architecture does not have a single point of failure and is resilient by design.
Automated Lineage Tracking and Caching
Lessen the execution time
Backtrack to the error-source
Cache repeated actions across multiple executions. Track all the generated data and visualize the data dependencies.
Great for platform builders
- Use built-in platform extensions like AWS Sagemaker, Distributed Pytorch, Distributed TensorFlow, AWS Batch, Kubernetes Pods, Containers, or use an extension of your own.
- Add a Python extension by simply extending a single class.
- Add a new language SDK.
Facilitate Team Collaboration
Various domains’ expertise onto one platform
Flyte makes it easy to stitch together workflows from different teams and domain experts. It simplifies the complexity of multi-step, multi-owner workflows.
Build Self-Service Data Platform
Break down data silos
Empower teams to self-service their data and ML pipelines while maintaining a centralized infrastructure platform that enables secure governance.