Every year, America's school districts burn through more than $8 billion keeping the lights on and the heat running. That's according to EPA figures last updated in 2025—and it's a number that facilities managers will tell you is probably understated. The real waste lives in the details: thermostats running at full blast over long weekends, lighting schedules no one has updated since 2018, billing errors that slip past finance departments month after month.
Which makes it prime territory for the latest crop of AI-powered energy optimization startups. The newest entry comes from two founders who, combined, don't yet have four decades of life experience between them.
Edviro—accepted into Y Combinator's Summer 2026 batch—is the work of Hursh Shah, 18, and Tanuj Siripurapu, 19. Their pitch: "energy world models" that catch waste before it shows up as a line item shock in next quarter's budget. The company reports identifying and verifying more than $400,000 in savings across a pilot deployment involving seven schools—though that figure remains unverified by third parties. Whether that number represents cumulative potential, annualized savings, or actual dollars clawed back from utility companies isn't entirely clear from public materials.
But it's gotten them into YC, which means $500,000 in funding ($125,000 for 7% equity plus $375,000 on an uncapped MFN SAFE) and a crash course in turning a technical concept into a scalable business. Now comes the harder part.
World Models Meet Aging School Infrastructure
The terminology Edviro uses—"world models"—comes straight from AI research labs, where it typically describes learned simulations of how systems behave over time. Apply that to buildings, the founders argue, and you get something more sophisticated than simple rule-based automation.
Here's how they describe the mechanics: The platform pulls 15-minute interval data from utility meters, runs anomaly detection against expected consumption patterns, then simulates what happens if you adjust setpoints or shift schedules. The system is structured around three phases—Identify, Delegate, Verify—though the naming feels a bit too tidy for the messy reality of building operations.
In the Identify phase, the software flags deviations and models potential interventions. Delegate is where things get interesting, or potentially concerning depending on your comfort with autonomous building control: AI agents either make approved adjustments automatically (thermostat tweaks, lighting schedules) or generate work orders for human staff when the fix requires someone with a wrench. Verify closes the loop, comparing results against actual utility bills to confirm the savings materialized.
The use cases span the usual suspects: HVAC optimization, lighting controls, demand charge management, tariff error detection. Edviro's website also mentions augmented reality features for maintenance tasks and service history logs, though no one outside the company appears to have seen those in action yet.
A Single Pilot and a Lot of Extrapolation
Both founders arrived at Edviro with atypical résumés for teenagers. Shah lists authorship on a NeurIPS 2024 paper, work with OpenAI's "ChatGPT Lab," and a pending medtech patent. Siripurapu cites experience at RTX and claims to have scaled a digital agency while still in high school. Before Edviro, they collaborated on Skyglass, which got them into the PearX accelerator.
The savings figures they're citing come from one district partner: seven school sites, 16 meters monitored. Siripurapu's resume, last updated perhaps three months ago, echoes the $400,000 figure and notes the platform caught utility overbilling among other inefficiencies.
What's conspicuously absent is external validation. No customer names. No third-party case studies. No detailed methodology for how those savings were measured and verified—a gap that procurement officers steeped in IPMVP standards (the industry benchmark for quantifying energy savings) will notice immediately.
It's company-reported data, in other words. And while that doesn't mean it's wrong, it does mean it's unverified in the way that matters to skeptical buyers with public budgets.
Stepping Into Territory Already Claimed
The building energy optimization space isn't exactly greenfield. The big incumbents—Siemens, Honeywell, Schneider Electric—have all grafted AI-driven features onto their building management platforms over the past couple of years. Siemens' Building X now includes AI-based energy modules with ISO 50000 certification. Honeywell rolled out an AI-powered building management solution around mid-2025.
Then there are the AI-native players, some of which have been at this long enough to accumulate actual proof points. BrainBox AI has deployed autonomous HVAC optimization across thousands of buildings and can point to case studies: 13% savings across a 600-store Dollar Tree rollout, 16.7% electricity reductions in partnership with Trane. Verdigris pairs energy AI with proprietary meter hardware. Clockworks Analytics specializes in fault detection and continuous commissioning. Cortex and Facilio go after enterprise office towers and commercial portfolios.
Edviro's focus on K-12 districts does offer some differentiation, perhaps more than the founders initially realized. Schools typically don't have enterprise IT budgets or in-house energy engineers. Companies like EnergyCAP already serve this segment with billing analytics and baseline modeling. But whether a two-person team can carve out defensible market share in that niche—against both well-funded startups and building management giants increasingly interested in the same customers—is an open question.
The Roadmap Forward (or Not)
What Edviro needs now is straightforward, if not exactly easy: move beyond pilot metrics that only the founders can vouch for. Facilities managers making purchasing decisions will want customer logos they can call. Climate tech investors will want granular savings breakdowns with dates, methodology, and ideally some third-party validation. The absence of public technical documentation, product demos, or IPMVP-compliant M&V reports leaves noticeable gaps for a platform whose central promise is verified savings.
YC's half-million-dollar check gives Shah and Siripurapu breathing room to scale the pilot and hire beyond their current duo. If the energy world models deliver as advertised—and if the verification methodology holds up under the kind of scrutiny that comes with selling to public institutions—there's a real opening here. School districts are hungry for accessible, automated energy management. They just need to believe it works.
If it doesn't, or if Edviro can't produce the evidence quickly enough, the company risks joining the long list of climate tech pitches that sound compelling in accelerator demo days but fade under competitive pressure and the unforgiving arithmetic of enterprise sales cycles.
For now, the founders are 18 and 19, funded, and chasing an $8 billion problem. That's not nothing. Whether it's enough is another matter entirely.
