01
Multi-model forecasting
Comparison of weather models, forecasts by location and tools to make sense of uncertainty.
In production
Snowy: a weather platform with maps, real-time data and AI.
Snowy is a weather platform for forecasts, maps, stations, alerts, reservoirs, air quality, earthquakes and smart tools, in a fast experience aimed at real decisions.
The difference is bringing product, data, SEO, interactive maps and agents together on an architecture built for it: SSR frontend, backend as the source of truth, per-domain cache and dedicated services for radar, CMS and batch jobs. Over the last 3 months Snowy has passed 14.7 million impressions, 242,000 organic clicks and 1,400 registered users.
- Role
- Design, development and infrastructure
- Period
- Since 2025, ongoing
- Scope
- Whole product, from database to SEO
- Reach
- 14.7M impressions in 90 days
01The underlying problem
Sixteen models that disagree.
Each source gives a different forecast for the same point, and the further ahead you look, the further apart they drift. The product work is not showing them all: it is giving one answer and saying how much it can be trusted.
Spread between modelsThe answer
Now+3 days+10 days
A schematic of the behaviour, not a measurement: what it shows is that uncertainty grows with lead time.
02Product
What Snowy is
A weather platform for Spain: multi-model forecasting over live data, with interactive maps and an assistant that answers in plain language.
02
Weather map
Radar, stations, alerts, earthquakes, air quality and environmental layers in one interactive interface.
03
Live stations
A station network with current and historical data, plus an onboarding flow for users.
04
SEO content
Indexable pages for locations, phenomena, pollen, air quality, reservoirs and WikiMeteo.
05
AI assistant
Conversation, voice, specialised tools and answers grounded in weather data.
06
Derived products
Energy, embeddable widgets, an eclipse product and new verticals on the same technical base.
Modules
Different surfaces, one architecture.
Each module has different requirements: external data, cache, visualisation, SEO, real-time state and shared internal models.
AI assistant
Conversational queries with weather tooling: clothing, alerts, forecast, location and actionable answers.
Reservoirs
Official data, weekly evolution, maps and comparisons by region, province and river basin.
Historical climate
Processing of historical series to analyse trends, anomalies and temperature change by area.
Earthquakes
Real-time monitor with official sources, magnitude, location, event detail and community reports.
Stations
Station detail with current metrics, history, favourites, owner and live weather data.
What shows up in usage.
SEO, performance and product usefulness already show up in usage: organic search, clicks and registered users on a platform of my own.
14.7M
impressions
last 3 months in organic search
242k
clicks
traffic from Google over 3 months
1,400+
registered users
an owned base for community and new features
16
models
ECMWF, GFS, ICON, ARPEGE, GEM and more
1,000+
stations
official network and Snowy community
370+
reservoirs
status and evolution across Spain
1,000+
terms
WikiMeteo in Spanish
20+
AI tools
assistant, voice and daily decisions
10+
map layers
radar, stations, risks and air
03Engineering
How it is built.
Snowy runs on a decoupled architecture: Next.js for SSR, SEO and UI; NestJS for business logic and data; Redis for cache; MySQL for persistence; and separate services where radar, CMS or jobs carry different loads.
FrontNext.jsServer rendering, SEO and interface
EngineNestJSBusiness logic, integrations and data model
CacheRedisWhat gets asked a lot and changes little
PersistenceMySQLThe state that has to survive
Separate
- Radar
- CMS
- Jobs
They split off when their load looks nothing like the rest: the radar renders tiles, the jobs run on a schedule.
Front
Next.js · React · TypeScript · Tailwind · MapLibre
Backend
NestJS · Prisma · MySQL · Redis
Infrastructure
Docker · Coolify · Cloudflare · Own VPS
AI and data
AI SDK · RAG · Batch jobs · GRIB2
Architecture
A multi-repo ecosystem with a Next.js front, NestJS engine, Vite CMS, Node.js radar, batch jobs and cross-cutting documentation.
Technical SEO
Indexable pages for cities, models, tools, stations, reservoirs, earthquakes, air quality, pollen, alerts, WikiMeteo and special content.
Data
Professional models, stations, reservoirs, earthquakes, air quality, pollen and official sources unified under one internal model.
Infrastructure
Production on VPS, Docker, Caddy, Cloudflare, GHCR, GitHub Actions, healthchecks, rollback and operational runbooks.
Radar
Interactive map with radar, stations, earthquakes, air quality, risk zones and environmental layers in real time.
AI
A weather assistant with natural language, voice mode and specialised tools that turn weather data into practical decisions.
B2B
Snowy Energy, embeddable widgets and sector verticals as a natural extension of the core weather product.
04SEO and data
SEO, data and performance as architecture decisions.
Users and Google both need fast answers. That's why the project runs on SSR, per-domain cache, an internal data model, IndexNow, revalidation and provider abstraction.
Integrated sources
AEMET
Stations and warnings
Euskalmet
Basque Country stations
MeteoGalicia
Galicia stations
MITECO
Reservoirs and water reserve
IGN
Earthquakes in Spain
USGS
Earthquakes worldwide
CAMS
Air quality and pollen
The goal is to unify heterogeneous providers into one consistent model, precompute the expensive parts and answer the end user very fast.
05B2B inside Snowy
Energy, widgets and sector verticals.
The same data, maps, forecasts and AI make derived products possible: energy forecasting, embeddable widgets and tools for specific cases.
Snowy Energy
Renewable forecasting, simulator and dashboard for solar energy as a B2B vertical inside the Snowy ecosystem.
snowy.es/productos/energiaB2B widgets
An embeddable SDK to bring weather data, maps, tools or the AI assistant into third-party sites.
snowy.es/productos/widgetsEclipses
A content and planning product: it launched with the total eclipse of August 2026 and already runs for the 2027 one.
snowy.es/eclipse-202706Press
Snowy has had public reach too.
The project was born out of LaRiojaMeteo and has appeared in press, radio and public portals. That's a signal of a real product, a community and continuity.
RNE
A radio interview to explain it
Radio appearance explaining Snowy and how the weather project has evolved.
larioja.com
Coverage in the regional press
Regional press coverage of the Snowy launch out of LaRiojaMeteo.
datos.gob.es
Public listing
Snowy is listed on Spain's national open data portal.
El Confidencial
18 January 2026
La Rioja Meteo launches Snowy, an AI-powered site for everyday weather decisions.
larioja.com
22 January 2026
Snowy is born: a La Rioja assistant for umbrellas, laundry and daily recommendations.
eldiario.es
12 January 2026
La Rioja Meteo creates Snowy, a weather platform with practical recommendations.
nuevecuatrouno
12 January 2026
Snowy launches as La Rioja Meteo's new weather platform.
Diario de León
14 April 2026
Snowy as a tool for planning eclipse viewing in León.
Actualidad Rioja Baja
12 January 2026
La Rioja Meteo launches Snowy as an advanced, accessible weather platform.
nuevecuatrouno
11 April 2026
Snowy and La Rioja Meteo in the planning for August's eclipse.
Does any of this fit what you need?
I work remotely, on European hours. If you have something in mind, tell me and I will say honestly whether I am the right person.