When the lights go out, everyone suddenly notices the complex systems powering our cities. Residents of the San Francisco Bay Area experienced this firsthand during scheduled power outages implemented to prevent devastating wildfires. These precautionary measures followed catastrophic fires caused by faulty equipment, particularly transformers, highlighting the critical need for advanced maintenance solutions.
Transformers serve as vital connections between power plants, transmission lines, and distribution networks. A single transformer failure can plunge entire power plants into darkness, forcing operators to work tirelessly assessing components, analyzing diverse data sources, and determining necessary repairs or replacements. This traditional reactive approach often proves costly and inefficient.
The financial impact of power equipment failures extends far beyond lost revenue. Businesses halt operations, people find themselves trapped in elevators and subways, and educational institutions shut their doors. The ripple effects of these failures underscore the importance of proactive maintenance strategies powered by artificial intelligence.
Enter Tagup, an innovative startup revolutionizing transformer and industrial equipment maintenance through cutting-edge AI technology. Their comprehensive platform enables operators to consolidate all data streams into a unified dashboard while leveraging sophisticated machine learning algorithms to predict component failures before they occur.
Co-founded by CEO Jon Garrity '11 and CTO Will Vega-Brown '11, SM '13—who recently completed his PhD in MIT's Department of Mechanical Engineering—Tagup currently monitors approximately 60,000 pieces of equipment across North America and Europe. Their technology safeguards critical infrastructure including transformers, offshore wind turbines, and water filtration systems.
"Our mission is to harness artificial intelligence to make the machines powering our world safer, more reliable, and more efficient," Garrity explains, highlighting the company's commitment to transforming industrial maintenance through AI innovation.
The Spark of Innovation
Vega-Brown and Garrity's paths crossed multiple times during their MIT years. As undergraduates, they shared courses while Vega-Brown pursued dual majors in mechanical engineering and physics, and Garrity studied economics and physics. Their connection extended beyond academics to fraternity brotherhood and football teamwork.
Garrity's entrepreneurial journey began during MIT's Energy Ventures class and continued at the Martin Trust Center for Entrepreneurship. Later, while Garrity attended Harvard Business School and Vega-Brown pursued his doctorate, they reunited in MIT's New Enterprises course, planting seeds for their future collaboration.
The entrepreneurial spark ignited in 2015, after Garrity's tenure at GE Energy and during Vega-Brown's PhD research at MIT's Computer Science and Artificial Intelligence Laboratory. At GE, Garrity observed an innovative business model where critical assets like jet engines were leased rather than sold, with manufacturers maintaining responsibility for remote monitoring and maintenance.
"When I worked at GE, I constantly wondered: Why isn't this service available for all equipment types? The answer lies in economics," Garrity reflects. "Establishing remote monitoring centers, instrumenting field equipment, staffing engineering experts, and providing customer support requires substantial investment. Only when equipment failure costs—both in business interruption and repair—outweigh these high fixed costs does the model become viable."
"We recognized two crucial insights," Garrity continues. "First, with increasing sensor availability and cloud infrastructure, we could dramatically reduce monitoring costs. Second, emerging machine learning techniques could significantly enhance the productivity of engineers who manually review equipment data."
These insights birthed Tagup, though proving their technology required time and perseverance. "The challenge with applying AI to industrial applications lies in the scarcity of high-quality data," Vega-Brown explains. "While many customers possess massive datasets, industrial data often contains low information density. This necessitates extremely careful signal detection and model validation to deliver reliable forecasts and predictions."
The founders leveraged their MIT connections to launch their venture. They received guidance from MIT's Venture Mentoring Service and participated in the first cohort of the MIT Industrial Liaison Program's (ILP) STEX 25 accelerator, which connects promising startups with industry leaders. Through ILP, Tagup secured initial customers who helped train and validate their machine learning models.
Enhancing Power Reliability Through AI
Tagup's platform consolidates all customer equipment data into a single, sortable master list displaying each asset's disruption probability. Users can click specific assets to view historical data charts and trends feeding into Tagup's predictive models.
Rather than deploying proprietary sensors, Tagup integrates customers' real-time sensor measurements with additional data sources like maintenance records and machine parameters to enhance their machine learning algorithms' accuracy.
The founders adopted a focused approach to system development, beginning with transformers before gradually expanding to other asset categories.
Tagup's first deployment in August 2016 was at a power plant near MIT's campus along the Charles River. Just months later, while Garrity was overseas, he received an urgent call about an unexpectedly offline transformer. Using his phone, Garrity accessed real-time transformer data, providing the plant manager with critical information to restart the system. This intervention prevented approximately 26 hours of downtime and saved $150,000 in potential revenue.
"These incidents represent catastrophic business outcomes," Garrity emphasizes, noting that transformer failures alone cost an estimated $23 billion annually.
Since then, Tagup has secured partnerships with major utility companies including National Grid and Consolidated Edison Company of New York.
Looking ahead, Garrity and Vega-Brown are excited about applying machine learning to equipment operation control. They envision systems that can self-regulate, similar to autonomous vehicles detecting and avoiding obstacles.
These capabilities promise to transform the systems ensuring reliable power when we flip our switches at night.
"The truly exciting frontier is optimization," Garrity states. Vega-Brown concurs, adding, "Vast quantities of power and water are wasted simply because there aren't enough experts to tune controllers on every industrial machine globally. By capturing expert knowledge in AI algorithms, we can dramatically reduce inefficiency and enhance safety at scale."
