Gather AI, a startup that offers an AI platform for warehouse cameras and drones, has raised a $40 million Series B funding round led by Smith Point Capital. That\u2019s the VC firm founded by former Salesforce co-CEO Keith Block.\u00a0<\/p>\n
The Gather team first met Smith Point a year ago at a logistics conference, and \u201cit took Keith and his team five minutes to get what we\u2019re doing,\u201d co-founder and CEO Sankalp Arora told TechCrunch.\u00a0<\/p>\n
What Gather AI<\/a> is doing is unusual. The three founders met as PhD students at Carnegie Mellon University, where they built one of the first autonomous helicopters and tested it on the FBI training grounds in Quantico. (Block is a trustee for CMU.)\u00a0<\/p>\n In 2017, the founders took what they learned about teaching helicopters to fly and land safely and launched Gather AI. Using off-the-shelf cameras placed on strategic moving equipment like forklifts, as well as off-the-shelf drones flying around the warehouse, the cameras watch on-the-floor operations and log what they find into the warehouse management systems.\u00a0\u00a0<\/p>\n But the catch is, the AI isn\u2019t being random about what it scans. It is being \u201ccurious,\u201d as Arora described it.<\/p>\n \u201cMy PhD work focused on how to make different kinds of flying robots curious,\u201d he said. \u201cSo they\u2019re curious about boxes and bar codes and workflows.\u201d\u00a0<\/p>\n In addition to barcodes, they look for lot codes, text, expiration dates, case counts, damages, occupancy, and other items. The idea is that they will discover and predict issues like low inventory, misplaced stock, and workflows that may cause safety issues.\u00a0<\/p>\n Techcrunch event<\/p>\n \n\t\t\t\t\t\t\t\t\tBoston, MA They also work in environments unfriendly to people, like freezers and cold storage.\u00a0<\/p>\n Because Gather\u2019s underlying tech was built years before the age of large language models, this is not the kind of AI that an LLM uses.\u00a0<\/p>\n \u201cThey\u2019re not end-to-end neural networks,\u201d Arora explains. \u201cThey are classical Bayesian techniques, combined with neural networks.\u201d\u00a0\u00a0<\/p>\n AI vision Bayesian techniques use probability-base methods to teach computers how to interpret visual data. These systems allow the technology to learn by using data and prior knowledge to make decisions \u2014 meaning they don\u2019t suffer the hallucination problems of LLMs.\u00a0<\/p>\n Instead they \u201cget curious,\u201d as Arora\u00a0put it, to gather information (hence the startup\u2019s name) and make a decision on the next action based on what they\u2019ve learned.\u00a0<\/p>\n As old-school as that sounds, Gather AI is sitting at the edge of the next big thing in AI, sometimes called \u201cembodied AI.\u201d<\/a> These are robots that interact with the real world, as opposed to an LLM interacting via computer chat or web app.\u00a0\u00a0<\/p>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|
\n\t\t\t\t\t\t\t\t\t\t\t\t\tJune 23, 2026\n\t\t\t\t\t\t\t<\/p>\n