Two skiers built the world's best snow forecasting service and beat government weather agencies

Two ski enthusiasts, neither of them climate scientists or tech founders, built a snow forecasting app that outperformed US government services and major commercial brands. Their story shows how passion for the subject and custom AI models can win where big budgets fall short.

Where the accuracy comes from

The startup's founders spent years on the mountain long before they started writing code. Their approach is fundamentally different from what Weather.com or the government-run NWS offer.

The startup draws on three sources. The first is open data from government weather agencies: NOAA, the National Weather Service, satellite datasets, and readings from mountain weather stations. This data is free and available to anyone, but most apps use it as-is, without adapting it to mountain terrain.

The second layer is the team's own AI models, trained on the specific characteristics of individual mountain areas. Every ski zone has its own microclimate: how clouds interact with a ridge, how wind redistributes snow, which slopes get precipitation first. You can't look these patterns up in a reference book. They build up over years of systematic on-site observation.

The third element is the founders' decades of personal experience. People who have spent thousands of days on the slopes know that a forecast of "20 centimeters" can mean dream powder or an icy crust, depending on the overnight temperature and wind direction. That context is built directly into the algorithm.

Why the big players lose

Government services offer broad coverage but no mountain specialization. Major commercial brands have resources, but for them snow is just one of hundreds of weather categories. The result: skiers get an averaged forecast for the nearest town, not data for a specific ridge and slope aspect.

• Standard apps don't account for micro-terrain or slope orientation
• Government reports cover broad zones, not individual ski spots
• Major brands don't invest in niche mountain expertise
• The independent team updates its models based on live feedback from skiers
• The founders have skin in the game: they're heading to the mountain themselves next weekend

"The best forecast is one you trust enough to change your route," goes

the philosophy behind every algorithmic decision.

For the ski community, a difference in accuracy isn't an abstract percentage. It's a decision: whether to drive three hours to the resort, whether to pick the north or south slope, whether to head out early in the morning or after lunch. A bad forecast costs time and money.

The next frontier: avalanches

The team is working on avalanche danger forecasts, a logical and vital extension. Avalanches kill several dozen people every year in the US alone, and most victims are experienced freeriders who made decisions based on generalized data.

Avalanche forecasting requires exactly the skills the startup already excels at: granular data on snowpack layers, precipitation history over the preceding several days, overnight temperature swings, and local microclimate. Government avalanche centers work with broad zones; the team is aiming for pinpoint forecasts for specific routes and passes.

What this means

Niche AI startups with real domain expertise are increasingly beating general-purpose players. Everyone starts with the same raw data; the difference is who knows how to interpret it and why. A small team with a personal stake in the outcome builds a product that a corporation with a thousand employees can't replicate simply by throwing more budget at it.