Stanford founders using AI to fix India’s power grid

In May 2024, residents of the high-rise Supertech Ecovillage 2 in Greater Noida took to the streets late at night to protest against ongoing power outages. The country was reeling under record heat, with temperatures touching 50°C, and families described “immense hardship due to frequent power outages and inadequate supply.” The episode underscores a growing national risk: as heatwaves intensify and demand surges, India’s grid is being pushed to the edge.

That problem is the focus of a new startup, Pravāh, an AI grid intelligence startup now backed by top-tier investors including Khosla Ventures, Pear VC and Conviction. The company was started by co-founders Mohak Mangal, Dhruv Suri, and Aman Gupta on the Stanford campus. They decided to focus their energies on one of India’s most critical problems and gave it a Sanskrit name: Pravāh or “flow.”




Their timing is critical. India’s power sector stands at a crossroads with growing demand. However, the country’s discoms (state-run distribution companies) operate with forecasting tools built for a different century. Many rely on spreadsheets and simple statistical models, adequate on ordinary days but deeply blind to a late monsoon or to a burst of rooftop solar.

When forecasts fail, outcomes are painful: outages and, at times frantic, expensive purchases on the spot market, costs that eventually find their way into government deficits or the monthly electricity bill of a family.

“Indian discoms cannot afford to operate with yesterday’s information, at a time when energy independence is critical and the discoms are a huge burden to the taxpayer”, said Mohak Mangal, Pravāh’s co-founder.

Pravāh’s answer is an “AI native” decision support engine that is built specifically for this new environment. The company combines machine learning models for demand forecasting with high-resolution weather modelling that pinpoint how much purchases discoms should make and how they can reduce their AT&C losses. The company is already working with companies in the US, Germany, and India.

A crucial layer of the company’s ambition rests on weather itself. In a recent paper, the team with a Stanford professor, Aditi Sheshadri, examined top AI weather models against real Indian observations and found an unsettling pattern: many systems underestimate extreme rainfall and misjudge cyclone paths. In a monsoon country, those errors can be existential. These findings now inform Pravāh’s India-specific forecasting architecture, built on the premise that global models rarely understand local weather dynamics.

Dhruv Suri, who recently completed his PhD from Stanford in Energy Science, says, “Machine learning architectures give operators something they have never had before: foresight into the next hour, the next day, and the next extreme event. That is the difference between reacting to a crisis and preventing one.”

The founding team behind Pravāh brings an unusual blend of deep technical expertise. Mohak has worked with the World Bank and MIT’s Poverty Action Lab across India and Myanmar. Dhruv has published research on renewable integration, built real-time grid tools at Google X, and co-founded a social enterprise delivering clean energy to rural communities. Aman, with a PhD from NYU Courant and research stints at LMU Munich and Stanford, has co-developed climate-foundation models with NASA and IBM and has written widely on monsoon dynamics.

For the founders, their mission is about preventing exactly what families in Greater Noida endured that night. If India can forecast its next thunderstorm and peak electricity demand with precision, then no parent in Supertech Ecovillage 2 will have to carry their children into the streets at midnight just to demand the most basic right in a modern economy: reliable electricity.
Select The Times of India as your preferred source on Google Search