See the wind before it arrives.

Freyvind gives ski jumping officials a 30–60 second wind forecast, reducing delays, improving safety, and keeping broadcasts on the air.

Freyvind — from Old Norse: freyr (fair, blessed) + vindr (wind)

WIND FORECAST TIMELINE

now +15s +30s +45s +60s

How It Works Live Demo Get In Touch

The Problem

Ski jumping jury directors control the green light based on instantaneous wind readings. There is zero ability to see what's coming.

📡

Reactive, Not Predictive

Up to 12 ultrasonic anemometers measure wind along the landing hill during each jump. But nobody measures what's heading toward the hill in the next 60 seconds.

📺

Lost Broadcast Revenue

Wind holds create dead air during premium TV windows. Every minute of delay costs networks money and loses viewers.

⚠️

Athlete Safety Risk

Starts are blocked when wind exceeds 4 m/s or turbulence spikes above 2 m/s — but only after it happens. A gust that arrives mid-flight can't be compensated for with math after the fact.

⏱️

Competition Flow

The jury's only forward-looking tool is a 2-minute rolling average — not a forecast. Long wind holds freeze athletes on the inrun, disrupt momentum, and frustrate spectators.

How Freyvind Works

Upstream sensors detect wind conditions before they reach the hill. Machine learning translates that into a real-time forecast for the jury.

Upstream Sensor Array

3–5 ultrasonic anemometer stations placed at 250m intervals along the prevailing wind corridor, up to 1km from the hill.

Site-Specific ML Model

Trained on wind propagation patterns unique to each venue's terrain. Learns how air moves through the specific landscape surrounding the hill.

Jury Decision Support

A forecast overlay alongside the existing wind monitor: predicts when conditions will cross FIS blocking thresholds (4 m/s absolute, 3.5 m/s cross wind, 2 m/s turbulence) at 30, 60, and 90 second horizons.

⚡

No Cloud Dependency

Edge compute runs inference locally at the hill. Works offline during competition — no internet required.

📊

Broadcast Integration

Wind forecast data feeds directly into TV graphics, giving viewers insight into the jury's decision-making in real time.

🔧

Non-Invasive

Complements the existing EWOXX/Swiss Timing wind measurement system — doesn't replace it. Augments the jury director's authority. No FIS rule changes required.

The Team

Built by people who understand the problem from both sides of the green light.

Jessica Jerome

Sport & Safety Advisor

U.S. Olympic ski jumper (Sochi 2014). 13-time national champion. Active FIS official. She's waited for the green light at the top — and now makes the call from the other side.

Eric Muehlstein

Technology & Founder

Sensor networks, edge computing, and machine learning. Designs and operates LoRa mesh networks — field-deployed wireless sensor infrastructure in all conditions.

TBD

Venue Operations & Trial Partner

Venue access, operational expertise, and where innovation gets tested first.

Proof of Concept

Westby, Wisconsin — the Snowflake Ski Club, one of America's oldest ski jumping venues. 2026/2027 season.

SUMMER 2026

Instrumentation & Data Collection

Install upstream anemometer array along the prevailing wind corridor. Begin continuous data collection correlating upstream readings with hill-level conditions.

FALL 2026

Model Training & Validation

Train site-specific ML model on collected data. Validate prediction accuracy at 30/60/90 second horizons. Target: >80% accuracy at 30s.

WINTER 2026/2027

Live Competition Trial

Deploy jury decision support display in advisory mode during the Snowflake Tournament. System shadows jury decisions without overriding. Measure prediction accuracy and potential time saved.

SPRING 2027

Results & Partner Outreach

Compile trial results. Produce comparison video: "what the jury saw" vs. "what Freyvind predicted." Approach technology partners with proven concept.

Proven Science

The wind energy industry already predicts wind 10–60 seconds ahead using scanning Doppler LiDAR. We're adapting proven technology to a new domain.

🔬

Wind Energy Precedent

Look-ahead LiDAR systems at wind farms routinely achieve >85% prediction accuracy at 30–60 second horizons. The physics is proven.

🏔️

Shorter Distance, Easier Problem

Wind farms predict over 5–10km. We're predicting over 1km or less. The shorter distance partially compensates for mountain terrain complexity.

🎯

Lower Bar Than Energy

Wind farms need precise speed predictions. Ski jumping needs reliable gust detection — a fundamentally easier binary classification problem.