Precise avalanche forecasting for the backcountry.
Regional forecasts cover a whole mountain range and miss the storm pockets that loaded your aspect. Slope Labs reads the snowpack for the exact slope you're on, when you're planning the line, and when you're standing on top of it.
Research preview · Winter 26/27 · Founding cohort open
See the snowpack for a specific slope
Pick a point on the map and Slope Labs surfaces that location's layer profile, snow structure, hand hardness, and temperature, by aspect and elevation, refreshed hourly.
The bulletin is the authority.
Your regional avalanche center is the trusted source, and its bulletin is necessarily broad. A single danger rating can cover thousands of square kilometers and every aspect in it, and it holds until the next issue while the weather keeps moving. We sit alongside it at a finer grain.
The synthesis is on you.
Bulletin, SNOTEL, point forecasts, slope-angle maps, webcams. The data exists; the integration happens in your head, in the dark, before coffee.
Resolution has arrived.
Operational weather models now run at 2–3 km and refresh every hour. Snowpack physics runs in research labs. No one has yet brought them together for the person standing at the trailhead.
How it works
The short version of the methods section.
- 1
Weather in, every hour
Fresh runs of NOAA's HRRR (3 km) and Environment Canada's HRDPS (2.5 km) arrive every hour, supplying precipitation, wind, and temperature fields at the resolution terrain actually varies. Your inputs refresh through the day, not once before dawn.
- 2
Physics core
A physics-informed snowpack model simulates layer formation, settlement, and stress, constrained by real snow science rather than pattern-matching alone.
- 3
Learning layer
Graph neural networks propagate snowpack state across terrain, because what loaded the bowl above you matters to the slope you're on.
- 4
Synthesis, on your clock
We pull weather, snowpack physics, and terrain into sub-regional instability signals by aspect and elevation band , refreshed hourly, and up to every 15 minutes when conditions are moving fast. The integration that used to happen in your head before coffee is already done.
- 5
Explainable, not a black box
Every signal is traceable to the weather and snowpack drivers behind it, not a single opaque score. You can see why a slope is flagged, weigh it against your own observations, and bring it to your local bulletin.
Who is building this
Slope Labs was founded by Brandon Hurd, a backcountry traveler first and a machine-learning researcher second. By day, he builds embedded AI labs for enterprise clients; by obsession, he studies snow.
Today, we are taking the heavy computational architecture of physics-informed ML out of the lab and grounding it in backcountry reality. We aren't here to replace the regional bulletin or your own judgment. We're building a new lens for the 5 a.m. planning ritual, taking the resolution from the macro-regional level down to the sub-regional scale, and eventually to terrain-specific detail.
What this is not
Slope Labs is decision support, not a decision-maker. It does not replace the official avalanche bulletin, formal avalanche education, partner checks, or your own observations in the field.
No model, ours included, can tell you a slope is safe. We're building a better input to your judgment, not a substitute for it. Read your local bulletin first. Then read us.
Founding cohort · Winter 26/27
Tell us your home zone and what a season of hourly, sub-regional snowpack signals would be worth to you. Your answers decide where we build first.
FAQ
When does it launch?
A research preview opens for the founding cohort in winter 26/27, starting with a small number of zones chosen from cohort demand.
How is this different from the bulletin?
It isn't a replacement. It's a complement at finer resolution and a faster clock: forecast centers rate a region once a day, while we model at the sub-regional scale and update hourly. Same humility, smaller grid cells, fresher data.
Will it cost money?
Eventually, yes. But we're not taking money until the model has a full validation season behind it. Founding-cohort pricing will be set from the survey responses.
What happens to my data?
Your email and survey answers are used only to plan and open the research preview. No selling, no sharing, no spam.