A New Mexican who has helped build an innovative training center in Santa Fe has co-founded a venture studio-backed startup, with the goal of integrating artificial intelligence into the world of advanced manufacturing.
Sarah Boisvert, who founded and operates an advanced manufacturing training organization in Santa Fe called the New Collar Network, recently joined Fab.ai as a co-founder and advisor. Boisvert previously co-founded Baltimore-based Potomac Photonics Inc. and has over three decades of experience in the advanced manufacturing industry.
The startup is developing an artificial intelligence (AI) model to support advanced manufacturing, which encompasses the use of more high-tech technologies, like 3D printing and automation, in manufacturing processes.
Specifically, Boisvert said Fab.ai is building a proprietary "large language model," or LLM, a type of AI deep learning model that uses massive amounts of data to come up with responses to various questions.
How accurate a model's responses are depends on the data used to train the model. Boisvert said she and other advanced manufacturing experts have curated industry sources from where the model pulls its data.
Those data sources include materials not publicly accessible on the internet, like technical papers behind journal paywalls, she added.
"We're not just scraping the internet where there's tons of wrong information," Boisvert said. "We're really going for what is accurate."
One of the experts who's helped Boisvert is Scott Halliday, a faculty member at Navajo Technical University in Crownpoint. Halliday runs the university's Center for Advanced Manufacturing; Boisvert said Halliday is a "true metal 3D printing expert."
The end goal of the model, Boisvert said, is addressing issues she's observed in her decades of advanced manufacturing experience. Industry standardization and skilled workers are two issues Boisvert said she's seen the model could assist with.
Boisvert gave the example of a manufacturer on a factory floor who can't remember "what temperature setting you have for your powder-bed fusion build area." Instead of relying on manuals, which might not include specific information, or internet sources, which might not be accurate, that worker could ask Fab.ai's LLM the question and it would "very quickly" provide an answer, she explained.
"We see it as a mentor on your shoulder," Boisvert said.
The model could also assist in design processes, she added, by taking concepts from "prompt to product."
"I can say to the machine, 'Please design for me a gear with 32 teeth that has X dimensions,' and whatever parameters you need to include, and it will design it within a few minutes," Boisvert said.
She added the LLM could help bring design processes "from 20 hours down to 20 minutes, roughly."
But there's a small caveat.
"You have to talk to them nice," she added. "You have to talk to them like they're humans."
Fab.ai currently plans to offer its LLM via individual subscriptions to engineers or other company officials as a way to "evangelize" the product, or as an enterprise software product that entire entities could access, Boisvert said.
It's also raising money for a seed round. Boisvert said the startup doesn't yet know how much money it's targeting through that funding round.
To test its model, Fab.ai is using Fab Lab Santa Fe, an advanced manufacturing fabrication facility at the Santa Fe Higher Education Center operated by New Collar Network that's a part of the larger Fab Labs network under the Massachusetts Institute of Technology's Center for Bits and Atoms.
Fab.ai is one of the three startups spun out by Roadrunner Ventures Studios, a deep tech venture studio headquartered in Albuquerque's Innovation District. Roadrunner announced its first three portfolio companies, including Fab.ai, at an event in December 2023; the studio provided $100,000 to each company and assists with ongoing product development and other early stage business efforts.
Michael Howard is Fab.ai's CEO. The startup has a four-person founding team and an eight-person engineering team, Boisvert said.