Sema4.ai, a new artificial intelligence startup founded by a group of prominent software executives, today announced that it has raised $30.5 million in funding.

The capital was provided by Benchmark and Mayfield Fund. The investment reportedly comes a mere three months after Sema4.ai’s launch.

Sema4.ai is led by Chief Executive Rob Bearden (pictured), who previously held the same role at data management provider Cloudera Inc. He was earlier the CEO of Hortonworks Inc., a Hadoop distributor. Sema4.ai’s other co-founders include former Cloudera engineering executives Ram Venkatesh, Suds Menon and Paul Codding. 

The founding team also includes Antti Karjalainen, CEO of business task automation startup Robocorp Inc. In conjunction with today’s announcement of its $30.5 million raise, Sema4.ai disclosed that it has acquired Robocorp. The latter company previously raised more than $25 million in funding from Sema4.ai backer Benchmark and other investors.

Robocorp offers a cloud platform, Control Room, that allows developers to create automation workflows in the Python programming language. Those workflows can perform repetitive tasks such as moving data between applications to save time for a company’s employees. In a blog post today, Robocorp stated that its services will remain accessible to customers following the acquisition.

“Our founders each have a proven track record of building significant enterprise software businesses,” Bearden said. “Our acquisition of Robocorp was borne out of a deep conviction that intelligence without action is incomplete, and action without intelligence is, at best, nonstrategic.”

Sema4.ai is building software that helps enterprises automate repetitive manual work using large language models. The company is seeking to develop LLM-powered tools that can not only retrieve information and summarize it, but also perform actions in business applications.

Sema4.ai’s first offering is an open-source framework called AI Actions that is hosted on Robocorp’s GitHub page. Using AI Actions, developers can create software modules that allow an LLM to interact with external business applications. An application team could, for example, build a module that enables an LLM to run queries on a database or browse the web using a search engine.

Similarly to Robocorp’s commercial Control Room platform, AI Actions uses Python as its script development language. Python is widely used in machine learning projects because it has a relatively simple syntax that eases common coding tasks. Additionally, there are many prepackaged code components available for the language, which removes the need for developers to build their applications from scratch.

According to Robocorp, AI Actions automates the task of setting up an AI application’s Python environment. That’s the collection of software modules necessary to run Python code. The framework also includes observability features that helps developers track how well their AI software works. 

AI Actions can optionally be used together with another open-source framework called LangChain. The latter tool makes it possible to equip an LLM with retrieval augmented generation, or RAG, features. LangChain also simplifies a number of related tasks such as testing LLM-powered applications’ reliability before their release.

Photo: SiliconANGLE