Zoltan's FLOPs is a mini-grant program for GPU computing projects. The goal is to provide small-scale funding to help researchers and developers access GPU compute time on modern hardware.
Total grant budget this iteration: $5,000.
Rules
Anyone can apply by filling out a simple form. There is no geographical restriction. The deadline is the 15th of April, 2025. Decisions will be released by 15th of May.
1-5 projects will be selected, depending on their compute requirements. The selected projects will get a portion of the total grant budget and can spend it on Vast.ai to provision on-demand GPU instances.
Grantees must use their GPU credits by the 31st August, 2025 and submit a 1-2 paragraph summary of their results.
Grantees agree to explain me (the funder of the grant) their project in detail on a video call and agree that a short, 1-2 paragraph description of their project will be published on this website, along with the final summary to be published later. Vast.ai instances will remain accessible to me to prevent abuse (e.g. crypto mining). I'm happy to sign an NDA.
The selection criteria are purely based on my personal judgment, taking into account the project proposal, the background of the applicant, and the computing requirements. In other words, cool projects with reasonable compute needs by applicants without funding will be prioritized.
Besides the Vast.ai credits, no further computing resources will be granted. Grantees must ensure their code is tested and ready to run on the GPU instances. The GPU configuration will be chosen by the grantee from the given budget.
Publishing code or technical details is not required. The grantee retains all rights to their code and results.
Apply here. Applications closed. Watch this site for results and future iterations.
Why?
I grew up in an era when computing was fairly democratic.
A kid in a poor, post-communist country with a used 486 could pick up coding and create programs just like anyone with better means.
Today, computing is becoming less accessible to the masses.
AI and machine learning require expensive hardware and significant energy resources.
Even top universities struggle to keep up, falling behind for-profit organizations. Students and researchers find it difficult to get GPU time for their projects. Talented kids interested in AI may never get the compute they need.
Carbon footprint
To offset the impact of the GPU usage, 280kg of CO2 will be permanently removed using ClimateWorks. My estimate of the computational carbon footprint is based on [1].
Who is Zoltan?
I am a programmer & entrepreneur based in Barcelona, Spain. LinkedIn