Most teenagers who spot a billing mistake might flag it to a teacher, maybe collect some extra credit. Hursh Shah turned his into a business.
The error—a utility bill gone wildly wrong during a high school research project—became the seed for Edviro, a San Francisco startup that's now pitching what it calls "energy world models" for buildings. The company came out of Y Combinator's recent batch with backing from Reach Capital. Dealroom.co lists $125,000 from Y Combinator, though such figures for YC companies often reflect partial amounts from the standard batch investment and should be taken with a grain of salt. It's an origin story that sounds almost too neat, the kind venture capitalists love to tell. But Shah and his co-founder, Tanuj Siripurapu, are betting that their software can do something more ambitious than just catch billing errors: predict energy waste before it happens, simulate fixes in a digital twin, then verify the savings afterward.
Whether that's enough to break through in a market crowded with both legacy players and well-funded AI upstarts is another question entirely.
The Pitch: Digital Twins That Actually Close the Loop
Edviro's approach starts with a problem the Department of Energy has long quantified: roughly 30% of the energy consumed in commercial buildings goes to waste. Plenty of software companies promise to fix that. Most stop at monitoring dashboards and alerts.
Edviro's founders argue their platform goes further. The system ingests utility bills, interval meter data, work orders, and building management system exports to construct a continuously learning baseline of what "normal" energy consumption looks like for any given facility. When usage deviates, the software flags it—but instead of firing off an alert and calling it a day, the platform's world model simulates potential interventions. Test a fix in the digital twin first, then implement it in the real building.
For HVAC optimization in spaces like offices, the system taps into existing Wi-Fi access points or deploys its own occupancy sensors to match heating and cooling with actual room usage. It can raise work orders for facilities teams, guide technicians through repairs via a mobile app, or adjust setpoints and schedules autonomously—though always after testing changes in the model. The company describes its measurement and verification reports as "board-ready, audit-grade," referencing IPMVP standards familiar to anyone in the energy services world.
One case example on the company's site cites a 21.4% reduction versus baseline, though the example is mentioned without full context or verification details. It's the kind of detail a CFO might want before signing off.
Why Schools?
Edviro is starting with K-12 facilities, a deliberate wedge strategy. According to EPA ENERGY STAR data, American schools spend more than $8 billion annually on energy—a figure that hasn't budged much despite decades of efficiency initiatives. The company says it currently has seven school sites live, with 29 more in the pipeline, and claims $400,000 in savings for clients to date. Those numbers are self-reported and haven't been independently verified.
A recent company blog post illustrated the kind of portfolio-level insights the platform surfaces: in one unnamed district with seven schools, Edviro identified a single site consuming five times the electricity of comparable buildings. Another post walked through how the system decomposes utility bill line items to trace anomalies back to specific operational issues—a faulty boiler controller, say, or an HVAC schedule that was never updated after a renovation.
The company is also running a pilot with a construction firm for baselining and post-project measurement and verification. Early district partners are being offered free energy audits, according to posts from Reach Capital and Y Combinator published mid-year.
The Team and the Funding
Shah, who has a NeurIPS workshop paper to his credit and describes previous work at something called "ChatGPT Lab," teamed up with Siripurapu, who previously scaled a digital agency and worked at RTX. Y Combinator lists the founding team as two people, while LinkedIn suggests the company has somewhere between two and ten employees. (That range, in startup terms, usually means three or four.)
Dealroom.co lists $125,000 from Y Combinator, though such figures for YC companies often reflect partial amounts from the standard batch investment and should be taken with a grain of salt. Reach Capital announced its involvement via LinkedIn, but didn't share round size or valuation. For a company this early, that's typical. Less typical: the fact that Edviro is already claiming paying customers and documented savings, even if the scale remains small.
A Crowded Field Gets More Crowded
The building energy optimization market is not exactly suffering from a lack of players. Honeywell's Forge Energy Optimization already offers autonomous machine learning for facilities. PassiveLogic announced what it termed "Level-3 autonomy" for buildings this summer, leaning on physics-based world models. BrainBox AI has published multiple case studies showing HVAC energy reductions in the 15% range or higher. Verdigris launched an M&V Analyzer in early 2025 and picked up strategic investment from Southwire shortly after.
In the K-12 segment specifically, EnergyCAP provides utility data management for districts, while Cenergistic runs behavioral energy-savings programs across schools nationwide. Traditional energy service companies like Schneider Electric and Ameresco deliver programmatic retrofits with measurement and verification baked in.
Edviro's bet seems to hinge on something subtler than pure technical superiority: the idea that facilities managers need a system that closes the full loop—detection, simulation, action, verification—rather than tools that hand off responsibility after the alert fires. The digital twin for capital planning is another angle. Testing whether to replace or repair a boiler, modeling how rate changes ripple through costs, simulating what happens when you add capacity: questions facilities teams face constantly, often without good data to guide the decision.
Whether that's differentiated enough to carve out meaningful market share against both entrenched software vendors and flush AI competitors is the gamble. Perhaps more pointedly: whether schools, notoriously budget-constrained and slow to adopt new technology, will pay for it at scale.
What Comes Next
For now, Edviro has seven schools running its platform and a couple dozen more in expansion. The real test isn't getting pilots—that's relatively straightforward in a space where everyone agrees energy is being wasted. The test is whether those pilots convert into repeatable sales, whether the $400,000 in claimed savings becomes $4 million, then $40 million.
Shah found a billing error as a teenager and turned it into a thesis. Now he has to prove the thesis works when someone else is signing the checks.
