Preteckting Buses: AI Company Uses Tech to Help NYCT Buses

It was recently announced that the New York City Transit Bus Maintenance (NYCT) tested artificial intelligence technology from Preteckt and will use it to help prevent bus breakdowns. 

Sasha Kucharczyk, co-founder and COO at Preteckt, spoke with METRO about this latest partnership, how the technology works, and more.

Preteckt is an AI startup focused on providing a maintenance tool that increases vehicle availability, improves safety, reduces costs, and supports maintenance teams.

AI Origins

The partnership between Preteckt and NYCT began when the company applied to an innovation challenge posed by Transit Tech Lab in New York.

Over the last two years, NYCT has tested Preteckt’s technology, and the company's founder Ken Sills said the technology is "a souped-up version of the check engine light in your car."

Kucharczyk breaks down the goal of this AI technology.

“The premise of what we’re doing is focused on helping technicians better understand the maintenance of their vehicles,” said Kucharczyk.

Preteckt accomplishes this goal by using AI technology to analyze the data generated by a vehicle and point the technicians to what they should be focusing on. Preteckt technology was developed for heavy-duty trucks and has adapted that technology for transit.

Diving into Data

The data the company leverages is millions of sensor data from each bus per day of operation. While NYCT had a system in place to collect this specific data, Preteckt developed algorithms and tools to interact with that data and identify maintenance implications.

“We provide high-precision repair plans,” Kucharczyk adds. “These repair plans identify upcoming maintenance issues, or even current maintenance issues that are minor but will progress to something more significant.”

Preteckt demonstrated its technology on 326 buses. These buses sent hundreds of millions of pieces of information from sensors to the company’s servers, where the data was then analyzed. 

During the last six months of the demonstration, there were about 50 buses sent to depots with repair plans from Preteckt.

Recently, the NYCT Maintenance team presented to the MTA Board, and some of the findings were highlighted, including a 75% increase in maintenance productivity and material costs lowered by 24%.

Before working with New York City, Preteckt worked with the Memphis Area Transit Authority (MATA) and several other agencies across North America. Kucharczyk pointed out the differences between these operations. 

“With other agencies, we had third-party hardware installed to collect the millions of sensor data we require,” Kucharczyk said. “With NYCT, they were already collecting the data we needed and our AI technology was added to the NYCT software stack.”

This meant Preteckt had to repurpose the total product to use a different data strategy to provide a certain result. Preteckt localizes its algorithms to ensure differences in bus or truck technology, as well as city-to-city variations, are taken into consideration. A truck operating in NYC has similar operating characteristics compared to buses or trucks operating in Seattle, for example.

Facing the Challenges

There were challenges on the technical side as the company began working with NYCT, but there were also challenges on the operational end.

Preteckt had a cellular connection with MATA that provided the company with constant data feeds coming off of the bus. In New York City, they did not have data coming off of the bus when they turned to the depot. This led to Preteckt taking a look at the data at the end of the day. 

Due to this data difference, Preteckt had to think ahead.

“We were looking at data a day back, and so one of the big differences with NYCT was to create a useful prediction for them,” Kucharczyk said. “It had to be a minimum of 48 hours before they had to take action. Anything less than that, and they were unable to take action on it. So we had to change our predictions in how it was done to accommodate that, and we really only need to look about 24 hours out. As long as they can make that prediction before the bus is in service, you can avoid any service or maintenance issue.”

Preteckt knows that understanding the customer is the most important aspect as they worked with truck OEMs, truck fleets, and different cities and agencies.

“The technical aspects of the technology might be the same across customers but you have to spend time understanding their operations. How we provide our predictions might need to change in order to make it valuable to the customer based on how they and their teams do their jobs,” Kucharczyk said.

As it stands, it is a one-year agreement between Preteckt and NYCT.

“We hope that the relationship will turn into a multiyear relationship with NYCT. We hope to work with more of their feet and help them with the maintenance of more subsystems of their buses. We have learned so much from the NYCT and appreciate the opportunity to introduce this AI innovation to the largest bus fleet in North America,” Kucharczyk said.