YC-Backed Startup Bobidi Raises $5.8 Million to Build Global Bug Bounty Platform for AI Models

Company building a global bug bounty platform to help companies validate AI models 4x to 10x more efficiently

LOS GATOS, Calif., July 14, 2022 /PRNewswire/ -- Bobidi, an AI model test platform, today announced it has raised $5.8M in seed funding from Y Combinator, We Ventures, Hyundai Motors Group ZER01NE, Scrum Ventures, New Product Experimentation (NPE), Meta, Lotte Ventures, Atlas Pac Capital, and angel investors. Bobidi will use the proceeds to mainly hire talent who will run the global bug bounty community, build the customer insight experiences, and develop various core machine learning technologies. 

Artificial Intelligence (AI) has exploded into every industry, transforming the way companies connect with their customers.  However, in the rush to build, test, and distribute AI, businesses often lack the resources and time to fully validate their models to ensure there is no potential bias or harm. This leads to a significant loss in consumer trust and confidence, and it in turn damages a company's brand.  Moreover, by the time bias or flaws are discovered in the model, the damage is already irrevocable.

Founded in June, 2021 by Jeong-Suh Choi and Soohyun Bae who left their jobs at Facebook and Niantic, Bobidi is building a global bug bounty platform that helps companies validate AI models 4x to 10x more efficiently.  The company's solution enables customers to connect AI models to the global bug bounty community via API in a secure way.  The Bobidi community can also test models and find biases (i.e. the edge cases that the models miss), through a seamless process that reduces the time needed for validation from months to a couple of weeks, if not days. The platform is also more efficient because the datasets collected via the bug bounty are the novel data that the model needs.

"Companies employing AI technologies require vast amounts of data to train and improve their AI models. However, it was previously almost impossible to guarantee that an AI model would perform well in the real world before actually deploying it because of the threat of unknown unknowns. The training and testing data sets may not fully capture the complexity of the problem in the real world," said Michael Yan, Head of Technical Investments at Scrum Ventures. "With Bobidi's massive network of testers around the world, companies now have a way to quickly test and validate their AI models by leveraging the power of a global community."

Bobidi believes the best people to find the flaws of the AI model are those that the AI model is targeting. Participants in Bobidi's global community are also benefiting from using the Bobidi platform to earn rewards.  In many countries, participants are able to make $10 to $20 per hour - which is substantially above local wages.  On top of the rewards, participants will enjoy tricking the AI models (i.e. finding the edge cases in a creative way), teaming up with others, developing their own track records, and building their own identities in the community.

"The era of big data is ending and we're about to enter the new era of quality data," says Choi.  "In the new era, people are tasked to find the best dataset for helping the model and Bobidi's unique approach of leveraging the global community is already well received by the market."

Sunita Parasuraman, Head of New Product Experimentation Investments at Meta said, "We're thrilled to back the talented founders of Bobidi, who are helping companies better validate AI models with an innovative solution driven by people around the globe." 

Choi explained, "The word Bobidi is a part of the spell from Cinderella, 'Bibbidi Bobbidi Boo' that turned everything ordinary into something extraordinary, like a pumpkin into a carriage. Like the spell, we hope to magically turn everything around people into a valuable datapoint for AI companies to create something great for the world."

About Bobidi
Bobidi is an AI model test platform for AI companies to safely validate models before deploying them. They leverage their global community of people to test models and find biases, which makes the entire process 10x more efficient.

To learn more, please visit www.bobidi.com.

SOURCE Bobidi