Marcelo Fornet is a software engineer with 12 years of experience blending competitive-programming rigor and production-grade distributed systems work, currently at xAI after leading ML and code-generation efforts at NEAR AI. A two-time ICPC World Finalist and Latin American Champion, he brings deep algorithmic intuition to problems in reinforcement learning, program synthesis, and blockchain consensus. His open-source contributions span high-profile projects like NEAR Protocol and a Rust EVM implementation, where he focused on transaction flow, fork detection, precompiles, and test automation. He has led engineering teams at Aurora Labs and shipped critical backend and tooling improvements across crypto stacks while also contributing full-stack solutions for competitive programming tooling. Based in Madrid with a background as a university professor, Marcelo combines research-minded exploration with hands-on system-building—often surfacing subtle protocol corner cases that others miss. His Github motto, "One man's noise is another man's ciphertext," reflects a taste for turning complexity into reliable, auditable code.
12 years of coding experience
7 years of employment as a software developer
Bachelor's degree Mathematics and Computer Science, Bachelor's degree Mathematics and Computer Science at University of Havana
Contributions:56 reviews, 165 commits, 160 PRs in 3 years 1 month
Contributions summary:Marcelo primarily focused on implementing and testing functionality related to the `txflow` module within the NEAR Protocol reference client. Their contributions involved adding and modifying tests to check the correctness of transaction flow, specifically within the DAG data structure. They also implemented a skeleton misbehaviour reporter and worked towards fork detection.
Browser extension which parses competitive programming problems
Role in this project:
Full-stack Developer
Contributions:5 commits, 7 PRs, 3 comments in 1 year 10 months
Contributions summary:Marcelo primarily contributed to the development of the `competitive-companion` browser extension. Their work included updating parsers for competitive programming platforms like COJ and Codeforces to correctly extract contest information. They also refactored code to leverage asynchronous operations for fetching problems and added functionality to parse contests lacking problem pages. Furthermore, they fixed parsing for specific sites like Yandex and incorporated a new host.
kattisbrowserproblemsjavascriptuva
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.