These days, just about everyone is trying to get a piece of generated AI action. While the focus is still on model vendors like OpenAI, Anthropic, and Cohere, or big companies like Microsoft, Meta, Google, and Amazon, there are actually startups that are trying to tackle the generative AI problem in a variety of ways. There are many. of the method.
Fireworks.ai is one such startup. Although it doesn't have as much of a brand name as some other players, the company claims it boasts the largest open source model API with over 12,000 users. This kind of open source traction tends to attract investor attention, and the company has raised $25 million so far.
Lin Qiao, co-founder and CEO of Fireworks, said her company doesn't train basic models from scratch, but instead helps customers fine-tune other models to fit the specific needs of their business. I pointed out that there is. “It can be an off-the-shelf open source model, a model that we have tuned, or a model that you can tune yourself. All three types can be provided through our inference engine API,” she said. she told TechCrunch.
Being an API, developers can connect it to their applications, deploy models of their choice trained on their data, and add generative AI features such as questions very quickly. Qiao says it is fast, efficient, and produces high-quality results.
Another advantage of Firework's approach is that it allows companies to experiment with multiple models. This is important in a rapidly changing market. “Our philosophy here is that we want to give users the ability to iterate and experiment with multiple models and provide effective tools for injecting data into multiple models to test products.” she said.
Perhaps more importantly, we keep costs down by limiting the size of our model to between 7 billion and 13 billion tokens, compared to over 1 trillion in ChatGPT4. This limits the world of words that large language models can understand, but allows developers to create much smaller, more focused data designed to work with more narrow business use cases. This will allow you to concentrate on your set.
Qiao is uniquely qualified to build such systems, having previously worked at Meta, and will build a fast, scalable development engine that powers AI across all of Meta's products and services. I lead the AI platform development team with this goal. She leverages her knowledge gained from her time at Meta to create API-based APIs that any company can take advantage of, without requiring the resources of enterprise-level engineering of Meta's size. I was able to create a tool.
The company raised $25 million in 2022 led by Benchmark with participation from Sequoia Capital, Databricks, Snowflake and other angel investors. The latter two are particularly interesting strategic investors because they are both data storage tools, and Fireworks allows users to leverage that data.