Archetype raises $35M to automate sensor data analysis with AI – News

Archetype AI Inc., a startup that provides artificial intelligence agents optimized to process sensor data, has raised $35 million in funding.

IAG Capital Partners and Hitachi Ventures led the Series A investment. Archetype stated on Thursday that the funds were joined by Bezos Expeditions, the Amazon Industrial Innovation Fund, Samsung Ventures and several others. The company previously raised a $13 million seed round last year. 

Palo Alto, California-based Archetype provides an AI platform that industrial companies such as manufacturers use to automate day-to-day tasks. The platform ships with three AI agents. One monitors industrial equipment for malfunctions, while another checks whether employees adhere to company best practices while carrying out their work. The third agent scans industrial environments for potential hazards.

Archetype’s platform also enables customers to build custom agents. According to the company, the task requires only a few lines of code. Archetype provides software building blocks that ease tasks such as securing custom agents and providing with them access to production line telemetry.

The company’s platform is powered by a series of AI models called Newton. According to Archetype, the models can scan camera footage and sensor readings from industrial environments to identify events of interest. One of the newest models in the series, Newton TimeFusion, also enables users to ask natural language questions about the data.

Some equipment failures can only be detected by reviewing telemetry from multiple sensors side-by-side . A sudden spike in machine temperatures, for example, might only require technicians’ attention if it’s accompanied by pressure fluctuations. Correlating different sensors’ readings in that manner can be challenging for AI models.

One of the reasons behind the task’s complexity is that industrial sensors vary significantly in how they collect data. The measurements they gather often describe different physical phenomena and use different data formats.

According to Archetype, its Newton models tackle the challenge by encoding multiple sensors’ output into a single embedding space. An embedding space is a mathematical structure that represents different types of data in the same numeric format. That unified format enables Newton to correlate readings from sensors with varying designs.

Newton also lends itself to other tasks, notably AI agent training. Industrial AI agents are usually trained on historical sensor measurements, which often contain data gaps. Newton can fill those gaps with synthetic, or AI-generated, data. It can also analyze historical telemetry to forecast whether a machine is likely to encounter errors in the near future.

Archetype says that its platform is finding use in multiple sectors including the construction, manufacturing and telecommunications sectors. The company will use its new funding to extend the platform’s capabilities. It plans to start rolling out feature additions in the coming months.

Photo: Unsplash