Onepanel Enables Forking of AI Models

We had the opportunity to meet with Donald Scott, Co-founder and COO, and Rush Tehrani, CEO and Co-founder of Onepanel during the IT Press Tour in San Francisco. Onepanel provides multi-cloud workflow and infrastructure automation for AI. AI workflows are hard to deploy. Onepanel is Kubernetes-native and enables the creation of custom container templates with write once and deploy anywhere capabilities.

You might also like:  Airflow to Orchestrate Machine Learning Algorithms

This gives enterprises the ability to bring in cross-functional teams to deploy AI enterprise-wide. There is less reliance on IT and DevOps teams by enabling people to innovate together with distributed containers across the workflow. Users have the ability to version control models, code, and datasets. GitLab is built into the platform. One panel uses open source across the board for server and client. You can fork a pipeline and send it to a customer to plug and play. The entire workflow is reproducible across the enterprise and with clients. Fork-able AI pipelines make it easier to replicate and reconnect pipelines.

You can plug-in any tool you want and scale to innovate much quicker without spinning up infrastructure makes it easy to deploy customized pipelines. Onepanel is essentially GitHub for AI pipelines.

Use cases include microscopy ingesting millions of slides across microscopes around the cloud to create a marketplace on top of their cloud offering. They’re able to compartmentalize and sell scaling terabytes (TBs) of data with AI.

A mobile application builder for the department of transportation vehicles looking for damaged potholes and street signs uses an edge compute device that produces 450 TBs per device. This enables them to create heat maps of trouble spots and validate budgets.

Mobile computer vision is connecting billions of phones processing TBs of data to make recommendations for improvement for cellular coverage and service.

You can have automated machine learning access to compute infrastructure when and where it is needed. Onepanel has also created a pre-built container metered using Unreal Engine.

There are more than 2,000 users on the platform built by word-of-mouth over two years. Many of the models are on GitHub so you are able to get started quickly.

Onepanel makes it easy to spin up a pipeline and fork it to save time, repurpose and see the workflow you’ve built. You can access additional GPUs to see the results and then shut down as an extension of your local machine.

Further Reading

Intelligently Automate Machine Learning, Artificial Intelligence, and Data Science

Leveraging AI and Automation for Successful DevSecOps