Concurrent.ai is a platform (in early development) that will enable you to get started with machine learning in a rapid and evolvable way.
See concurrentai/concurrentai-infra to start experimenting with your own Concurrent.ai stack.
From a technical perspective, Concurrent.ai is a generalized manifestation of the Rendezvous architecture detailed by Ted Dunning and Ellen Friedman in Machine Learning Logistics.
For a brief overview of the Rendezvous architecture and its many benefits, see Rendezvous Architecture for Data Science in Production by Jan Teichmann – a highly recommended read!
Although there are many benefits that the Rendezvous architecture offers, one major drawback is how high the initial engineering effort is to implement it. With Concurrent.ai, implementing a Rendezvous architecture will be as simple as writing a few lines of JSON.
Concurrent.ai will extend the Rendezvous architecture concept beyond machine learning and into general business logic, allowing you to start with a simple, non-ML solution first and seamlessly iterate towards ML without rebuilding your application.
All of the benefits that come along with a Rendezvous architecture: auto-scaling, ability to validate model behavior and performance in production without impacting users, not being locked into a single ML framework, and more.
- Rendezvous API
- Model Enricher (implementation in progress)
- Model Executor
- Rendezvous Collector
- Analysis Collector (not yet implemented)
See the Concurrent.ai Roadmap project for an up-to-date roadmap.
Pull requests are welcome! Many details here are still being worked out – see the Contributing Guide to get started. Everyone contributing to Concurrent.ai repositories or engaging in discussion is expected to follow the Code of Conduct.
Licensed under the Apache License, Version 2.0.