NuMind, a Y Combinator Summer 2022 alum and the maker of a software tool to create custom NLP models, today launched out of private beta and announced its $3 million in seed funding. Using large language models (LLMs) and its proprietary “Interactive AI Development” learning paradigm, the NuMind solution allows all companies–even those with no expertise–to quickly create machine learning models to perform such tasks as analyzing sentiment, detecting topics, moderating content, and creating chatbots.
The $3 million seed investment was led by Flybridge. Additional funding was provided by Brazil-based big_bets, Carya Ventures, the Pioneer Fund, Velocity Ventures, France-based Sharpstone Capital as well as Y Combinator. NuMind’s initial customers are developing custom models for medical coding, legal report review, user feedback analysis, and matching resumes with job postings for a recruitment firm.
“We are living a revolution thanks to large language models such as GPT-4, but tools to take advantage of them are not mature yet. Tackling applications such as content moderation or sentiment analysis still requires ML experts and an intense labeling effort,” explained Etienne Bernard, CEO, NuMind. “Our goal at NuMind is to make this revolution accessible to all, starting with language understanding. It’s a game changer in terms of speed and efficiency for all the companies needing this type of AI.”
The NLP models designed by NuMind’s first customers were developed dramatically faster than by using traditional methods–up to a factor of 10. At the end of the project, companies own their models.
“The NuMind founders recognized how LLMs could be leveraged as part of a platform to drive textual understanding in an easy-to-use, accessible manner so developers could speed the creation and deployment of NLP models,” said Chip Hazard, co-founder and general partner, Flybridge. “Now organizations can maximize the potential of NLP, building better applications and improving experiences.”
“Given the depth of this insight and the breadth of the market opportunity, we were thrilled to invest and help them realize this vision,” Hazard continued.
Interactive AI Development
NuMind’s Interactive AI Development paradigm is similar to how people are taught–with an interactive process of questions, answers and explanations–but doing this between a person and the machine. This type of learning is more natural and efficient, eliminates labeling large volumes of data, and allows much faster creation of personalized models.
“Up to now there were two ways to teach computers: either by programming or by giving them many examples. I see LLMs as the key to unlock this third way that we call Interactive AI Development,” said Bernard.
An all-in-one solution, NuMind offers:
- Reduction of labeling effort, thanks to LLMs
- Automatic machine learning to determine the best models automatically
- Active learning to only label the most important documents
- An intuitive labeling interface
- A live performance report to quickly identify the strengths and weaknesses of models as the project progresses
- Multilingual models to avoid any need for translation
- Full ownership of your models so you can put them in production however you want
NuMind was founded by two cousins, Etienne Bernard and Samuel Bernard. Etienne is an MIT postdoc, former head of ML at Wolfram Research, and the author of “Introduction to Machine Learning.” Samuel is the former CTO of Make.org, a platform that galvanizes citizens to propose solutions to public issues. The NuMind team of seasoned professionals is split between Paris and Boston. Early customers are based in Brazil, France, and the United States.
NuMind’s solution is available as a monthly subscription. For more information, visit www.numind.ai.