Image credit: Monciti/Getty Images
The success of large-scale language models like GPT has sparked a developer frenzy to create AI-powered applications. But building AI services can be challenging, especially given the lack of skilled developers to keep up with the recent surge in demand.
Enter Chaoyu Yang, an early software engineer at data mega unicorn Databricks. Together with his co-founders, he built the AI development framework BentoML and just announced a seed funding round.
In an interview with TechCrunch, Yang explained that today’s AI services are often built on multiple machine learning models, making them complex to manage and operate. Many of the programmers entering this fray come from full-stack or application development backgrounds. This means they often lack the skills to build the necessary AI infrastructure, resulting in lengthy development processes.
For example, a demo AI app like Microsoft’s Visual ChatGPT, a chatbot upgrade that allows you to generate responses from both text and image prompts, takes at least 3-6 months to be production-ready. Yang said it could take months.
While tech giants like Microsoft enjoy the financial and human capital to train AI models and use them in the real world, small businesses, in Yang’s words, “get a lot out of AI. It collects “valuable data” that could be of great benefit, but builds an infrastructure for development that is “short on resources.” “
BentoML provides a high-level API that abstracts the infrastructure details required to run AI models on the cloud. We are in a camp of tools like SageMaker that want to smooth the way for AI service development. It’s a so-called AI application framework, a set of tools that makes it easy to build, ship, and scale AI applications, like the construction tool kit you use when building a house.
Specifically, BentoML targets data scientists who train AI models, DevOp engineers who manage the lifecycle, and developers who actually build applications on top of the models.
With BentoML, developers can make Visual ChatGPT scalable and cost-effective for production use in as little as two days, Yang said. Users are also using this framework to run art generator Stable Diffusion and open source LLM on the cloud.
Yang compared his company to Vercel, which focuses on serving front-end developers and has a final valuation of over $1 billion. BentoML aims to be his vercel of AI, he said.
Yang predicted that AI would eventually become more production-ready, but admitted that he did not expect the wave of AI applications to arrive so quickly. The founders expect AI app developers to make up more than 90% of platform users in the future.
“If you asked me a year ago, probably 90% of companies would have trained their own models. But you can perform well.” I’ve seen it before,” he said.
“Rather than focusing on model training, developers can just focus on fine-tuning and product engineering, but the lack of AI-focused developers has become a bottleneck in itself. increase.”
BentoML was open sourced in 2019, after which a self-hosted SaaS version was introduced to enterprise customers. The company has acquired users organically through its open-source community, with membership quadrupling to more than 3,000 in the past year, and South Korean social networking giants Line and Naver are among his early adopters. has become one.
Yang declined to disclose the size of the company’s revenue.
Investors are noticing BentoML’s traction in the developer community. The startup recently raised $9 million in a seed financing round led by DCM Ventures, which also includes Bow Capital. DCM general partner Hurst Lynn joined BentoML’s board after the round.
A booming AI market has been a boon for BentoML, Yang admitted, but the rapid changes in the industry have made it difficult for the team to balance short-term and long-term goals.
“We may have to build something that rides the current trend, but in the long term, of course, we want to have our own moat. It’s about how you strike a balance.”
