Wow, what an amazing year we’ve had! And the future only looks brighter. I'm so excited to tell you about Clarifai Community, our next major product launch here at Clarifai. It's something we've been thinking about for over five years, going back to when we were raising our Series B. At that time we set out on a multiple-year journey to put all the tools for building world-class AI in one place, and in doing so we’ve become an AI accelerator for solving the most valuable business problems.
Last week we announced our Series C and now I would like to give you more details about an essential part of computer vision for Clarifai.
Under the hood, the AI Lake is much more than a storage layer for all your inputs. The AI Lake is really incredible because you can manage all your AI resources in one place: images, text, audio, videos, annotations, the models you train, the workflows you assemble, and the concepts you want to recognize in the world. It's powered by the tools for model and workflow management as well as a knowledge graph of concepts that the community is creating.
To understand the power of this, it’s useful to break down model development into the 2 core components of building AI—the data and the algorithms.
All data that is stored can be automatically labeled in pre-processing workflows at the time of ingestion, or by humans with AI-assisted labeling. We help you label data 100x faster with humans only doing a fraction of the work. We provide streamlined user interfaces to optimize the human workflow. All that data gets indexed so that you can find it and manage it at scale. And you can snapshot your datasets so you can train on those datasets, evaluate models on those data sets, and make your AI reproducible. This allows you to really experiment and iterate as you build your AI for scale.
We’ve integrated state-of-the-art toolkits like Hugging Face, Detectron2, OpenMMLab so that all the best algorithms from the research community are at your fingertips so that you can integrate them into your AI apps and workflows.
But this is only the beginning. Clarifai shines because of its flexibility. With a click of a button you can train state-of-the-art models without having to tweak knobs or tune hyperparameters. You can even fine tune your already built models using transfer learning, which is a technique that leverages the context of what a model already understands and learn new concepts quickly. The way we’ve done this is so powerful because you can train models with just a few examples and still get high accuracy results.
And because data science teams are already working on their models, you can upload these into Clarifai and automatically get all of the benefits of Armada, which automates all of the ML Ops - you never have to worry about the MLOps or the DevOps again.
Now that you’ve built great AI. We want you to share it with the world. We already have multiple ways to collaborate:
1. By inviting collaborators
2. Creating teams within your organization
With the launch of Clarifai Community, users can now make their AI resources discoverable by the public, for anyone to use.
From early on, we conceptualized AI Apps as the main container to house all the AI resources for a project in one place. This allows you to package all your work in a single container so that you can share and collaborate with others in the community and your own organization. And most importantly, because your AI is containerized, it's all reproducible.