Stack AI’s co-founders, Antoni Rosinol and Bernardo Aceituno, were PhD students at MIT in 2022 when large language models started gaining prominence. By the end of that year, ChatGPT would be released, but they had already identified an issue: businesses were trying to integrate data with models without sufficient expertise. They aimed to address this problem.
After completing their degrees, they relocated to San Francisco and joined Y Combinator’s Winter 23 cohort. There, they established Stack and refined their concept. Today, the company has developed a low-code workflow automation tool to help businesses build AI-driven workflows, such as chatbots and AI assistants. To date, the company has raised $3 million.
“Our platform enables people to create workflows that integrate various tools. We emphasize connecting data sources and LLMs, as this allows for powerful workflow automations. We also provide numerous other tools and functions for automating complex business processes,” Aceituno shared. Although their product has only been available for six months, they already have over 200 customers.
The process involves dragging components onto a workflow canvas. This typically includes a data source like Google Drive and an LLM, along with other components such as a trigger or an action component to build the workflow, making it possible for customers to create generative AI programs with minimal coding. While the coding is not AI-driven, the tasks within the workflow often are and may require some manual coding to ensure smooth operation.
Some of their initial customers are from the healthcare industry. Aceituno notes that they must be cautious with applications involving doctors and patients, particularly when internal data sources may not be reliable or could have conflicting or outdated information.
In such cases, he emphasizes the importance of relying on the human expert, the doctor, to judge the quality of the answers. To further ensure accuracy, they include source citations in every answer, so healthcare professionals can verify the information before accepting it.
“That being said, it’s true that inputting poor-quality data will result in poor-quality citations. That’s why these assistants should not take over the process entirely,” he added.
Having transitioned directly from MIT to launching a startup, Rosinol believes that joining YC was instrumental in understanding the business aspects and refining their startup concept through customer interactions.
“We began with an initial version of this API, which was much more developer-focused. Initially, we had a few clients and aimed to use AI to automate RFP responses or sales. By collaborating with customers, it became evident that the main challenge was not in training a model, but in effectively querying and linking data sources to these language models.”
The company currently employs six people but is expanding its team by hiring engineers and sales and marketing professionals.
The $3 million investment was secured about a year ago. Investors include Gradient Ventures, Beat Ventures, and True Capital, along with LambdaLabs, Y Combinator, Soma Capital, and Epakon Capital.
