Rainbow Weather raises $5.5M seed to build real-time environmental intelligence for a climate-volatile world
Rainbow Weather, a next-gen climate tech startup specialising in hyper-localised short-term weather forecasting, announced it has raised $5.5M Seed funding round. Rainbow Weather has developed a real-time, AI-driven weather intelligence platform that fuses satellite imagery, radar, meteorological stations, and even smartphone sensor data to deliver highly precise, minute-level forecasts and severe-weather detection. The company now sells its data both through a consumer app and directly to enterprises and other weather providers, positioning itself as an infrastructure layer for a world of increasingly volatile climate conditions. The company was founded in 2021 by Belarusian serial entrepreneurs Yuriy Melnichek and Alexander Matveenko. Melnichek previously built AIMatter, a neural-network platform later acquired by Google, as well as video app Vochi (acquired by Pinterest) and fashion tech startup Wanna (acquired by Farfetch). Matveenko, meanwhile, founded AI mapping company MapData, which he sold to Mapbox in 2017.
I spoke to the co-founders to find out more.
From colleagues to the hailstorm that exposed the limits of traditional forecasting
Melnichek and Matveenko have known each other since childhood and have worked together previously. Throughout, they kept the shared ambition to eventually build a startup together as co-founders. Melnichek recalled, “Some time later, I approached Matveenko and said: “How about we finally do it?” They started thinking about something at the intersection of maps and weather."
For Melnichek, it was personal.
“I was living in Switzerland and hiking in the Alps, using an app called RainViewer, which I really liked. I was literally hiking behind a rain cloud. The app showed that the cloud would keep moving and I would stay dry. But then I suddenly felt something hitting my head. It was hail, about the size of grapes. The rain cloud reached the mountains, stopped, intensified, and started hailing.
The app didn’t account for elevation. It didn’t know there was a mountain, so it relied only on past observations, not on how the terrain would affect the cloud.”
That was the moment he realised: existing models don’t really understand how weather systems evolve in real time in complex geography. S
o the idea to use machine learning, elevation data, and multiple sources of real-time input to predict how rain and other weather phenomena actually develop, not just where they were historically, was born. How Rainbow builds forecasts from space, radar, and smartphones
Rainbow Weather’s core product is a hyper-accurate minute-by-minute weather forecasting app powered by AI and machine learning.
It delivers four-hour precipitation reports tailored to the exact moment a request is made.
For instance, if a user checks weather at 3:51 am, the app will provide precise predictions through 7:51 am. The platform updates every 10 minutes. It also offers spatial resolution down to a single square kilometre (0.62 mi²). All of these features set the product apart from major competitors, including The Weather Company (formerly owned by IBM) whose forecasts refresh every 15 minutes and extend up to 7 hours ahead.
The team started with rain prediction, and then expanded to other weather parameters. Initially, Rainbow was built as a consumer weather app with AI-based rain prediction. But the team soon discovered that the forecast's quality is limited by the quality of the input data.
“If you put garbage in, you get garbage out,” shared Melnichek.
“So we started investing heavily in data acquisition and data fusion. We work with satellite imagery, which captures cloud systems from above across multiple infrared and visible spectral bands. You can roughly detect where precipitation is happening, but it’s not very precise.
Then there are meteorological radars – the big round radomes you often see near airports. They work like microwaves: they emit radio waves that bounce back from water particles, so they detect rain extremely well, but only what is already falling, not what is forming behind the front.”
Rainbow.ai also gathers ground weather station data, and started collecting air-pressure data from smartphones.
“Modern phones have barometers, originally introduced to measure altitude changes for fitness tracking, like counting how many stairs you walk, explained Melnichek.
But these sensors are very accurate, and pressure changes are highly correlated with weather dynamics. Rainbow Weatehr currently ingests data from more than 1,000 meteorological radars worldwide, multiple satellites, ground stations, and mobile sensors. Each source has its own processing pipeline. One central system then blends the outputs using neural networks.
Why Rainbow replaces batch forecasting with continuous atmospheric streaming
I wanted to understand why real-time data was possible with such a large data set.
According to Melnichek:
“Crucially, we do not operate in batch mode but as a continuous stream: as soon as a new satellite frame, radar scan, or pressure update arrives, it is immediately ingested and processed.
Each data source has its own pipeline feeding our neural networks, which continuously blend these inputs to produce the most accurate real-time representation of the global atmospheric state, updated every few minutes.”
Beating legacy models on timing, not probabilities
Although there are a number of prominent players in the market, including AccuWeather, Apple Weather, and The Weather Company, Rainbow Weather asserts that the current forecasting methods are outdated.
“Many legacy forecasting providers rely on optical flow for short-term precipitation forecasting. That’s a fast but simplistic method that treats clouds as shapes in motion, without any understanding of atmospheric physics,” explained Matveenko.
“A second category of services uses large-scale mathematical models that do incorporate physical principles, but they’re so cumbersome and slow that they can’t respond quickly to real-time weather changes.”
Rainbow Weather, by contrast, uses advanced machine-learning models to merge a vast array of high-resolution data to generate predictions.
“Mixing heterogeneous data allows us to eliminate the typical errors inherent to each individual source. This, in turn, helps us to feed cleaner and more accurate data into our models and achieve a much more precise forecast. And thanks to the optimised performance of our AI models, we can make this forecast much faster than our competitors,” Matveenko added.
From rain app to real-time environmental intelligence
Specifically, Rainbow.ai focuses on timing: the exact start and end of precipitation events. “On the consumer side, we deliberately do not show probability distributions, because most people do not understand them correctly,” shared Melnichek.
“There was even a Stanford study showing that users misinterpret “30 per cent chance of rain” in completely different ways – some think it means 30 per cent of the hour will be rainy, others think it means it will rain in 30 per cent of the area.”
For B2B clients, the company, of course, provides full probabilistic fields, confidence levels, and uncertainty bands. “But our competitive advantage is very precise nowcasting – what will happen in the next minutes and hours, not in seven days.”
The company has also expanded into fire detection, a move inspired in part by the severe wildfires Melnichek has witnessed in Cyprus, where he lives.
“We realised that we already process satellite imagery for the whole planet every few minutes, so why not extend our pipeline to detect thermal anomalies and smoke patterns that indicate fires?” he says.
“We don’t need to build a completely new infrastructure — we just add another model on top of the same data streams. The philosophy is the same: fast detection, continuous monitoring, global coverage.
“With climate change, wildfires, heatwaves, and extreme weather are becoming more frequent. We want Rainbow to become a general real-time environmental intelligence system, not only a rain app.”
The ‘wow moment’ driving organic growth
As of today, Rainbow Weather has reached over 1 million installs and over 100,000 active users. The app's strongest growth driver is word of mouth.
According to Melnichek:
"When people experience that the rain starts exactly when the app says it will start, and stops exactly when it says it will stop, they remember it. They compare it with the default weather apps on their phones, which can sometimes show sunshine while it is already raining outside. This “wow moment” creates very strong recommendations to friends and family.”
Opening the black box to transparent analytics
The team also runs weatherindex.ai, an open-source tool that evaluates the accuracy of short-term precipitation predictions from providers like AccuWeather, Vaisala, and The Weather Company in real time. It pulls live data from public APIs and compares forecasts with verified airport weather reports using standard metrics such as accuracy and F-score (a measure of predictive performance).
Melnichek says the decision to build an open benchmarking platform was driven by a belief that weather forecasting, like AI, should be judged through transparent, comparable, and publicly verifiable performance metrics.
"In the AI world, we are used to open benchmarks. Everyone compares models — GPT, Claude, Gemini, open-source LLMs — and you can see how they perform on standard datasets.
In weather, it is very different. Many providers explicitly forbid you, in their license agreements, from comparing their data with competitors’ data. That felt very outdated and very defensive to us.”
So the company decided to do the opposite.
It built Weather Index to compare different weather providers against verified ground truth, mainly using airport meteorological stations. Airports have high-quality instruments and human meteorologists who validate observations, so this gives a very reliable reference.
“We buy data from different providers, run the same evaluation across all of them, and publish the results openly. You can see which provider is more accurate in which country and for which forecast horizon. For example, for one-hour rain timing, we perform best in much of Europe — the UK, France, Italy, Spain, Finland, the Baltics, and Turkey. In North America, The Weather Company performs extremely well. In Japan, AccuWeather is very strong. In Southeast Asia and Australia, we are again among the leaders. We believe weather data is critical infrastructure, and critical infrastructure should be evaluated transparently.”
The startup was backed by a syndicate of investors, including Yuri Gurski, founder and president of Flo Health, the first purely digital consumer women’s health app to achieve unicorn status. With the new funding, the startup plans to go beyond precipitation forecasts by incorporating additional weather parameters, extend the forecasting horizon from 4 hours to 24 hours, and grow its presence in the B2B segment of the weather-forecasting industry, which is projected to reach $4.07 billion by 2030. In the long term, Rainbow AI wants to build the most accurate real-time environmental intelligence platform in the world. Not just a weather forecast, but a continuously updating digital representation of what is happening in the atmosphere and on the ground – a system that helps people and businesses stay safe and make better decisions in an increasingly unstable climate.
Read More