Google Summer of Code 2025
At Open Climate Fix, encouraging open source contributions, building community, and working to make datasets more accessible is a key part of our mission.
This year we were lucky enough to be granted the opportunity to mentor across six projects as part of Google Summer of Code, a mentoring programme run by Google.
Want to learn more? We’ve broken down each project below, as well as a bit about our mentees.
Open Data PVNet
We're building an open-source solar forecasting pipeline to integrate with OCF's PVNet model, using publicly available Numerical Weather Prediction data to forecast solar generation at the national level, starting with the UK. Currently, our main forecasting tool, Quartz Solar, is trained using a mixture of public and private datasets, and we want to create an effective model that uses 100% open data.
Using our readily available data our GSoC mentee, Siddharth, is implementing machine learning to train the model. By the end of the project, we are hoping to have a UK-based ML solar forecast trained on free NWP data, with the accuracy benchmarked.
Meet the Mentee: Siddharth
I applied to OCF for GSoC because I believe open-source climate tools can create real impact, especially in underserved regions. With Open Data PVNet, we're making high-quality solar forecasting accessible by using only publicly available weather datasets. No paywalls, no black boxes — just open science that anyone can build on."
Explore the project
Testing the Streaming Capability of Icechunk on Cloud Storage
We use a large amount of Satellite and Numerical Weather Prediction data all saved in Zarr format when training our ML models. We normally have a local copy, rather than using the cloud. This project explores using Icechunk. We're planning to create a benchmark when training our PVNet model with data in cloud storage (using modern stack like Icechunk + Zarr 3).
We hope to then use this to measure speed and compare speeds when training using data on disk.
Meet the Mentee: Dakshbir
I chose this project with Open Climate Fix for two reasons. First, the mission is incredibly compelling. The chance to apply my software engineering skills to accelerate climate research was a powerful motivator. Second, it presented a fascinating technical challenge. OCF works with petabyte-scale climate datasets which was new and exciting for me."
Explore the project
Cloudcasting UI
We've been working on forecasting clouds using satellite imagery, and we would now like to take the next step and make an exciting visual representation of these cloud forecasts. This is an exciting and innovative project in the weather forecasting space. We would either implement this into our current product, Quartz Solar Forecast, or a separate tool all together.
Meet the Mentee: Suvan
Why OCF? I wanted to work with an organization that uses tech for good, something that not only challenges me technically but also helps tackle climate change. OCF stood out because it's doing exactly that, using open source and AI to make renewable energy work better. Their work in regions like Rajasthan, India (my second home <3) made it feel even more personal. I was genuinely excited to be part of that mission and contribute to their amazing work."
Explore the project
Quartz Solar: New data source in ML model
Adding new data sources usually gives a boost to the predictive power of our models, and finding innovative ways of extracting information from them can be even more beneficial. This project explores ways to improve our solar energy forecast with an ablation study of how much data on things like dust or neighbouring sites can contribute to the precision of the model.
Meet the Mentee: Zaryab
I applied to Open Climate Fix because I was very interested in learning how large scale machine learning models work. This specific project will teach me all aspects of a ML project, everything from data gathering and transformation to training models and doing analysis."
Explore the project
Quartz Solar: New data source in ML model
Quantile Regression in probabilistic solar forecasting directly estimates specific, fixed quantiles of the predictive distribution. This provides discrete output points, but inherently limits the comprehensive capture of complex uncertainties.
The proposition of this project involves a fundamental modification to the final model layers. Instead of fixed quantities, our mentee will aim to predict the parameters of Gaussian Mixture Models (GMMs) – specifically the mean (μ), standard deviation (σ), and mixing coefficient (α) for each component.
The expected outcome is a complete, continuous probability distribution of solar power.
Meet the Mentee: Tara
I applied to OCF because I wanted to work with a company that’s clearly trying to make a real impact, working on climate prediction, and tackling the tough challenge of translating academic research into something usable in industry. As a PhD student, I see this gap all the time, and I wanted to help bridge it. I chose this project specifically because my PhD is on probabilistic machine learning methods for quantum computing hardware, and I wanted to apply those transferable skills to a field I care about. OCF was the perfect opportunity for this, and my project, which involves applying probabilistic ML methods to OCF’s models, grabbed my attention straight away!"
Explore the project
TabPFN as a replacement for the adjuster
For the OCF Quartz Solar forecasting model, we have a simple “adjuster” model. It currently looks at the pattern of errors in the last week, and adjusts the new forecast. This project experiments with a new foundational time series forecasting model, TabPFN, with the planned outcome to dynamically adjust our PV forecasting model based on historical errors and improve the forecasting skill.
Meet the Mentee: Anshul
I applied to GSoC with OCF because I’m passionate about sustainable AI and eager to explore energy-aware solar forecasting. Through this project, I’ve been able to contribute to open-source research at the intersection of efficient computing and machine learning. My work focuses on evaluating the performance and trade-offs of TabPFN—a lightweight foundation model for tabular data, and other ML models—against existing rule-based adjuster used in OCF’s PV forecast adjustment. The goal is to better understand how such ML models can power faster, greener, and more scalable real world forecasting applications."
Explore the project