AirOps is helping companies build AI-enabled applications on top of LLMs

There is a shift in the air, and it feels like companies need to be thinking about how to put large language models to work, but as with any new advanced technology, it’s often easier said than done, especially for less technical organizations.

AirOps, an early stage startup is in the right place at the right time, helping companies take advantage of these new capabilities to build AI-enabled applications on top of large language models. Today, the company announced a $7 million seed round, which actually closed at the beginning of last year.

Company CEO and co-founder Alex Halliday says that with the recent interest in LLMs, there is a challenge for businesses trying to get involved. “There is a really large gap to close between these amazing capabilities that folks can play with in things like ChatGPT, and then [applying that] to the kind of hardest challenges in the business. So we’re creating a platform that lets folks come in and create custom solutions on top of these algorithms that really move numbers in the business,” Halliday told TechCrunch.

The company is currently helping customers build applications on top of three LLMs: GPT-4, GPT-3 and Claude. The idea is to help users do things like automating processes, extracting insights from data, generating personalized content and performing natural language processing techniques. according to the company.

Halliday says that current customers are looking for ways to take advantage of their own data and content in conjunction with LLMs to build new content from that existing corpus or build a generative AI experience on top of their existing software.

One of the primary value propositions of the company is helping customers use these models more efficiently and effectively because it can get expensive. “What’s kind of really interesting is that you can actually use the larger models to train smaller models. So maybe for the first couple of months you would run using GPT-4, and that would create the training outputs to then use a smaller, open source model that’s been fine tuned,” he said.

And AirOps can help you move through those steps. “We’re really learning the right recipes and architectures here, but we expect over time the kind of boil-the-ocean, sledgehammer approach will give way to a more nuanced and a better understanding of how to take advantage of the menu of choices people have,” he said.

The company launched last year with the goal of helping get value from their organizational data, but as LLMs moved into the public consciousness, the company shifted its focus. “As we started to look at the application of LLMs to the data space, we realized that actually a much larger opportunity was helping people blend LLMs with their data to create custom workflows and applications,” he said. Last fall they really shifted their focus to that approach.

The company has 14 employees with a few open roles. Halliday says he sees diversity across many dimensions, but he is aiming to build a diverse employee base as he builds the company, and this is especially true given how new LLMs are. “We’ve really been very open minded when hiring to people with different backgrounds and different levels of experience,” he said.

The $7 million seed investment was led by Wing VC with participation from Founder Collective, XFund, Village Global, Apollo Projects and Lachy Groom.