.svg){kind=small}
.svg)
.svg)
WHOUMPF: AI Decision Support for Mountain Professionals
A sovereign AI sentinel designed to clear the "information blizzard" and optimize safety for alpine guides and rescuers.
In the high mountains, safety depends on clarity. Yet, every morning, mountain professionals (guides, rescuers, instructors) are hit by an "information blizzard." Between the BERA (Avalanche Risk Bulletin), weather models, wind history, and slope maps, synthesizing data takes 45 minutes of intense mental effort. This leads to decision fatigue before the first step is even taken on the snow.Project WHOUMPF (Weather & Hazard Observation Unit for Mountain Professionals Flow) is a sovereign AI-driven sentinel designed for the FFME community. Using Make, the system harvests official French data (Météo-France, BERA) every night. An AI layer, hosted in Europe, then analyzes 48-hour trends to isolate "red flag" anomalies,such as hidden weak layers or rapid warming,that the human eye might miss under pressure.As an industrial engineer and aspiring mountain guide, I built WHOUMPF on a strict principle: AI is a compass, not a pilot. It never replaces the "ground truth"—the sound of a real "Whoumpf" under a ski or the physical analysis of the snowpack. It simply clears the mental fog so the expert can focus on the terrain.Climate Resilience: By enabling early "No-Go" decisions at 5:00 AM, we prevent unnecessary car travel, saving roughly 1 ton of CO2eq per club season.Biodiversity: The tool integrates wildlife quiet zones (e.g., Grand Tétras) to ensure a respectful, low-impact practice.WHOUMPF proves that AI can be humble, local, and essential for protecting both people and our fragile alpine ecosystems.
In the mountains, information is a double-edged sword: lack of it is dangerous, but an overload of it is paralyzing. This project addresses three critical pain points faced by mountain professionals and FFME club leaders:1. The "Information Whiteout" (Cognitive Overload)Every morning before dawn, a guide must synthesize a massive volume of raw data: the BERA (Avalanche Risk Bulletin), wind speed/direction history, temperature gradients, and slope maps. This manual "data-crunching" takes 45–60 minutes. It creates decision fatigue—a dangerous mental state where the expert enters high-risk terrain already cognitively exhausted.2. The Hidden "Red Flags" (Detection Gap)Subtle but lethal patterns—like a rapid temperature spike combined with a specific wind shift—can be buried in dense PDF reports. In the rush of early morning preparation, even a seasoned expert can fall victim to confirmation bias, overlooking a "signal" that indicates a weak snow layer. WHOUMPF acts as a second pair of eyes, flagging these anomalies objectively.3. The "Last-Minute" Logistics Trap (Environmental Impact)Without a clear synthesis at 5:00 AM, groups often drive 1.5 hours to the trailhead only to find the conditions too dangerous to climb. This leads to frustration and unnecessary CO2 emissions. By providing a "Strategic Brief" before the car even leaves the garage, we enable smarter "Go/No-Go" decisions, saving time, fuel, and reducing the human footprint on fragile ecosystems.WHOUMPF doesn't solve the mountain's mysteries—it solves the human struggle to process them. It clears the "administrative fog" so the professional can arrive at the trailhead with a fresh mind, ready to focus on the only thing that matters: the actual, physical reality of the terrain.
The WHOUMPF solution is engineered for robustness, sovereignty, and low-bandwidth reliability, utilizing a "Smart Ridge" architecture built on a No-Code backbone.1. Sovereign Data Ingestion (The Basecamp)The system uses Make (hosted on European servers) to orchestrate the workflow. Every morning at 4:00 AM, a scheduled trigger executes a multi-source harvest:API Integration: It pulls raw data from Météo-France (Open Data), specifically targeting massifs in the French Pyrenees.OCR & Parsing: Since the BERA (Avalanche Bulletin) is often a dense PDF, the system uses a parsing module to extract semantic data regarding snow stability, wind-loading, and elevation-specific risks.2. Intelligence Layer (The Compass)The extracted data is fed into a Large Language Model (GPT-4o/Claude 3.5) via a specialized System Prompt.Chain-of-Thought Analysis: The AI doesn't just "read" the data; it compares the last 48 hours of weather history to identify deltas (e.g., “Rapid warming (+5°C) since midnight on North-East slopes”).Signal Detection: It is programmed to isolate "Red Flags"—patterns that historically lead to "Whoumpf" sounds or slab releases—and cross-references them with IGN slope maps (>30°).3. Lean Distribution (The Radio Link)Airtable (EU Region): Acts as the central database, logging every brief to build a "Memory of the Mountain" for long-term safety audits.Mobile Briefing: The output is a formatted Markdown brief pushed via Telegram API. This ensures high legibility on small screens and successful delivery even with a weak 4G/Edge signal at the trailhead.By using Make, the solution remains transparent and reversible. As an industrial engineer, I designed it so that if a data source fails, the system immediately alerts the user: "Source unavailable—rely exclusively on manual field observation." It is a tool built for the rugged reality of the terrain.
The ROI of Project WHOUMPF is measured by the optimization of human cognitive resources and the reduction of environmental impact. For a mountain professional or an FFME club, the results are broken down into three strategic KPIs:1. 95% Reduction in Data Processing TimeBefore: A leader spends 45 to 60 minutes manually aggregating data (reading PDFs, checking wind sensors, comparing historical charts).After: The AI-synthesized Strategic Brief is delivered in under 2 minutes.Impact: This saves approximately 15 to 20 hours per month for a club running 4 weekly outings. This "reclaimed time" is directly reinvested into physical gear checks and human safety briefings with the group.2. Logistics & Carbon Efficiency (The "Smart No-Go")KPI: Percentage of "Dead-End Travels" avoided.Result: By identifying critical hazards at 5:00 AM rather than at the trailhead, the project prevents unnecessary vehicle trips.Quantifiable Impact: Avoiding just 10 group trips (3-hour average round trip) prevents the emission of approximately 1 ton of CO2eq per season. This aligns the club's activities with national decarbonization goals.3. Safety & Decision AccuracySignal Detection: In early testing, the AI flagged "hidden" temperature inversions or rapid wind-shift patterns in 15% of bulletins that were initially overlooked during quick human reads.Adoption Rate: 100% of the pilot group reported a significant decrease in "morning stress" and "mental fog," which are leading factors in human-error-related mountain accidents.4. Cost-to-Value RatioInfrastructure: Built on a "Lean" No-Code stack (Make + Airtable), the operating cost is less than €30/month.ROI: For a negligible cost, the community gains a high-performance safety sentinel that offers the same level of data synthesis as proprietary professional systems costing thousands of euros.
I should win this award because WHOUMPF is the perfect embodiment of "AI for Good"—a project where industrial engineering rigor meets high-altitude passion to solve a life-stakes problem.1. A Unique Bridge Between Two WorldsWinning this award would recognize a "hybrid" expertise. I have applied the same precision used in industrial process optimization to the unpredictable environment of the French Pyrenees. By mastering the "Product Builder" stack (Make, AI, No-Code), I’ve proven that a solo developer can build a sovereign, professional-grade safety tool that serves a prestigious national federation like the FFME.2. Innovation Rooted in Sobriété (Sustainability)This isn't technology for technology's sake. WHOUMPF is a model of "Low-Tech AI." It is:Carbon-Conscious: It directly reduces greenhouse gas emissions by optimizing mountain logistics and preventing "useless" car trips.Environmentally Ethical: It protects fragile species, like the Grand Tétras, by integrating biodiversity "Quiet Zones" into safety briefs.Sovereign: Hosted in Europe, using local data, it respects data privacy and territorial integrity.3. The "Human-in-the-Loop" PhilosophyMy project stands out because it refuses the "black box" approach. As an aspiring guide, I designed WHOUMPF to empower the human expert, not replace them. It acknowledges that while AI can process data at lightning speed, only the human eye can "feel" the snow.Winning would allow me to scale this "Sentinel" to other mountain clubs and massifs, proving that AI can be a humble, silent partner in protecting both human lives and our fragile wilderness. WHOUMPF is a project with a soul, a mascot (my Samoyed), and a mission: to make sure the beauty of the summit never comes at an unnecessary cost.
.svg){kind=icon}