With $1.3 million raised in seed money, Boston-based Tristar AI Inc., a new software development firm, is introducing "the supervisor that never sleeps" to the plastics industry.
The technology uses camera footage and algorithms to analyze human behaviors, then it alerts plant operators to maintenance needs, missed tasks and production interruptions.
"This is a new type of AI that helps supervisors keep track of what's happening on the factory floor," Tristar AI CEO Salem Karani said in a phone interview.
The 29-year-old Harvard University graduate founded Tristar AI in July 2022 with alumni of Massachusetts Institute of Technology, which he also attended. Its technology captures real-time information about plant performance using computer vision and AI to analyze human kinetics and imagery.
The devices have smart cameras to continuously monitor each production line, sending alerts of problems and data about production activities and frequency to supervisors. The goal is to minimize interruptions.
Karani said early adopters of Tristar AI technology include film extruders and injection molders using it for preventative maintenance and quality control related to scrap, vacuum system filters and pallet-wrapping cycle times.
"Now you know if the filter has actually been changed, was done incorrectly or not at all. The point is to help train your staff and avoid costly errors," Karani said.
Some injection molders are buying the technology to make sure molds are being lubricated correctly and daily.
The system provides video footage of anything that goes wrong.
"We're helping workers be more effective. You can train staff and show them this is actually how it's done," Karani said.
Leominster, Mass.-based United Solutions Inc., which molds Rubbermaid-brand consumer storage totes, is among the first customers. Pulp-and-paper company Georgia Pacific LLC in Atlanta is testing a pilot system.
Karani expects the next applications to improve plant safety for all processes and performance for thermoformers and blow molders.
"If there are manual tasks, you can get close to 100 percent accurate information about how well they are being done and how fast they are being done," Karani said.