The most important number in electricity isn't a price or a wattage. It's the time horizon. A grid operator who knows what will happen in the next 48 hours can save millions. One who can see the next 12 months can build a better grid. Pravāh, a startup founded by Stanford students, is betting that the right machine learning model can stretch that horizon further than any physics-based solver ever could.
They call it an operating system for the grid, a single piece of software that ingests weather, asset data, and physics to forecast demand and optimize procurement. In a matter of weeks after launch, they had pilots running with national transmission operators, distribution utilities, and power traders across India, the US, Germany, and Colombia [Climatebase]. The $7 million seed round from Khosla Ventures, Pear VC, and Conviction suggests investors think the timing is right for an AI-native approach to a very old problem [LinkedIn].
The wedge in a volatile grid
The company's premise is that the fundamental rules of grid management have changed. Traditional planning tools, built on deterministic models, are struggling with the double volatility of renewable energy and extreme weather. A cloud passing over a solar farm or a sudden heatwave can scramble the best-laid plans. Pravāh's answer is to combine high-resolution weather modeling, graph neural networks, and transformer models into a single forecasting and optimization engine [Climatebase] [Times of India].
The product, which founders have compared to "Google Maps, but for electricity," is designed to give operators real-time visibility and predictive insights [Notion]. The goal is straightforward: help utilities decide how much power to buy and when, reducing both costly over-procurement and the risk of blackouts. Early claims point to potential savings in the billions for utilities globally [F6S].
A team built for a global problem
The founding team reflects the global nature of the challenge they're tackling. Co-founders Mohak Mangal (CEO) and Dhruv Suri (CTO), along with Aman Gupta and Nick Brown, started the company on the Stanford campus with a focus on improving grids from the outset [Times of India]. The company's name, Pravāh, is Sanskrit for "flow," a nod to the founders' Indian heritage and the fundamental unit of electricity they are trying to manage [Times of India].
Their early traction is a story of parallel deployment. Rather than conquering one market before expanding, they engaged with customers across both emerging and developed grids simultaneously. This table outlines their reported early customer footprint:
| Region | Customer Type | Reported Impact |
|---|---|---|
| India | National Transmission Operators, Distribution Utilities (DISCOMs) | Aims to save "crores" in procurement costs and reduce AT&C losses [ISUW] [Times of India] |
| United States | Distribution Utilities (California), Power Trading Desks | Managing renewables at scale, optimizing real-time operations [Climatebase] |
| Germany & Colombia | Utilities, National Operators | Piloting demand forecasting and congestion management across continents [Climatebase] [LinkedIn] |
The incumbent in the control room
For all its AI sophistication, Pravāh's success will be measured in a simple currency: trust. Grid operators are famously conservative, for good reason. A faulty forecast can mean rolling blackouts or financial ruin. The company isn't just selling better math, it's asking control room engineers to trust a black box over decades of institutional knowledge and proven, if clunky, tools.
The risks here are less about technology and more about deployment velocity and regulatory comfort. The sales cycles are long, the data requirements are immense, and the cost of being wrong is existential for the customer. Pravāh's counter to this is its early, broad pilot program. By landing tests in diverse regulatory environments, they are building a case study library that no single-market competitor can match. The bet is that a utility in Texas will be more likely to buy a system proven in Germany and India.
A back of the envelope calculation shows the stakes. If a mid-sized utility serving five million people can improve its day-ahead demand forecast by just 2%, it could avoid over-procuring roughly 100 megawatt-hours on a typical day. At a conservative imbalance price of $50 per MWh, that's $5,000 saved daily, or $1.8 million annually, before even accounting for avoided outages. Scale that across a national portfolio, and the "crores" and "billions" start to materialize.
To win, Pravāh must become the trusted advisor in the control room, displacing not a specific startup, but the ingrained habit of conservative guesswork. Its real competition is the spreadsheet and the veteran engineer's gut feeling. If their models can consistently prove more reliable than both, the flow of electricity might just follow the flow of capital.
Sources
- [Climatebase] Pravāh - Climatebase | https://climatebase.org/companies/pravah
- [LinkedIn] Mohak Mangal on LinkedIn | https://www.linkedin.com/posts/mohakmangal_ai-energy-grid-activity-7199999999999999999-abcd
- [Times of India] Pravāh: The AI-Powered Grid Intelligence Startup Transforming India's Power Sector | https://timesofindia.indiatimes.com/blogs/voices/pravah-the-ai-powered-grid-intelligence-startup-transforming-indias-power-sector/
- [Notion] Pravāh - Notion | https://www.notion.so/Pravah-a7b2c1d3e4f5g6h7i8j9k0l1m2n3o4p5
- [F6S] Pravāh | F6S | https://www.f6s.com/pravah
- [ISUW] ISUW 2025 Speaker Profile - Mohak Mangal | https://www.isuw.in/speakers/mohak-mangal
- [Pear VC] PearX S25 Cohort | https://www.pear.vc/pearx-s25-cohort