A leading artificial general intelligence (AGI) research company focused on testing and validating a large number of projects from their research teams. This involved significant data labeling needs and their previous internal labeling tool took a substantial amount of time and engineering resources to develop and meet their expanding set of use cases. In the process of launching a large-scale AI-powered product to market, the company wanted to rapidly vet the content the AI application was generating, without running the risk of relying on a single labeling vendor for their projects. 

Locking into a single vendor typically meant cost overruns, delays and the inability to flexibly scale or switch as the project scope expanded. The company decided early on that it was necessary to maintain research project velocity by having the ability to manage multiple labeling vendors at once. To solve this bottleneck, the company leveraged Labelbox’s platform which provided the ability to test and contract multiple labeling vendors simultaneously. This allowed them to rapidly manage over 10 labeling vendors - from initial introduction to project kick off - and minimized the risk of not having a trusted vendor during this critical juncture.

As another key requirement, the company wanted to minimize any potential security incidents during the development of their AI product, which meant that any platform had to be vigorously vetted for both security and compliance. The Labelbox platform provided a robust infrastructure that provided enterprise-grade security to support and enable breakthroughs for their ML research and development.  

In terms of set up, the company only needed minimal technical resources in order to quickly test and validate projects within Labelbox. Their prior experience with other vendors was slower, and encountered significant delays kicking off labeling projects. With Labelbox however, the company was able to configure projects in a matter of minutes for their primary data types which included both images and text. In addition, the company utilized Labelbox’s webhooks and attachments to make the process of data import and offering labelers with contextual information about assets to be labeled, as easy as possible. Onboarding on to Labelbox required little additional engineering work and allowed the company to evolve their project details, ontologies and labeling workflows as their needs changed, depending on model performance and project trajectory.

By adopting Labelbox, the company was now able to review and generate hundreds of thousands of annotations in the span of just three months for their content moderation use case (which focused on tagging safe vs. unsafe content.) Expanding upon their trust and safety processes, the company integrated a real-time pipeline for flagging and reporting images with the ability to refresh their existing models. This accelerated the speed of removing unsafe content in the form of images that were violent, toxic, lewd, etc. The overall business value of being able to verify this content given the short timeframe was realized through a hugely successful product launch of their AI application, generating widespread industry excitement and user adoption.