Brooklyn AI Startup Lands MTA Pilot to Predict Subway Delays

NEW YORK — A three-year-old startup working out of a converted warehouse in Dumbo has won a six-month pilot with the Metropolitan Transportation Authority to test whether artificial intelligence can predict subway delays before they cascade across the system. The company, which employs a small team of transit engineers and data scientists, says its models ingest live signal data, train dwell times, and historical patterns to flag emerging disruptions up to forty minutes early.

The premise is that most major delays begin as small perturbations — a train held slightly too long at a busy interchange, a signal fault on a single track — that ripple outward until an entire line seizes up. By detecting those early signals, the company's founders argue, dispatchers could reroute trains or adjust spacing before riders ever notice a problem. "We're not trying to replace the humans in the control center," the chief executive said in an interview at the firm's office overlooking the East River. "We're trying to give them a forty-minute head start."

During the pilot, the system will run in an advisory capacity on two heavily trafficked lines, generating predictions that transit staff can compare against what actually unfolds. The MTA emphasized that no automated decisions will be made without human oversight, and that the pilot is strictly an evaluation. If the forecasts prove accurate, officials said, the tool could eventually inform real-time service adjustments and rider-facing alerts.

Independent transit researchers offered cautious interest, noting that delay prediction is notoriously difficult because the underlying causes are so varied and interdependent. "The subway is a chaotic system layered on top of century-old infrastructure," one researcher said. "A model that works in the morning rush might fall apart during a weekend track fire." She said the true test would be whether the predictions held up during the messy, unusual events that cause the worst delays.

The startup has raised a modest seed round from a mix of climate- and infrastructure-focused investors, and its founders framed the MTA pilot as a proving ground that could open doors to transit agencies in other cities. They were careful, though, to temper expectations, acknowledging that agency procurement moves slowly and that a successful pilot guarantees nothing. "Public transit is the hardest possible customer, and that's exactly why we wanted it," the chief executive said.

For riders, any payoff remains hypothetical for now. But the pilot reflects a broader push inside the transit authority to modernize how it manages an aging network, from new signaling projects to data-sharing partnerships with outside firms. The MTA said it would review the pilot's results in the winter and decide then whether to expand, extend, or shelve the experiment. "If it saves even a handful of the worst delays," one transit official said, "that's thousands of hours of people's lives back."