AlphaEvolve:
Agentic Optimizations of Algorithms & Code

About the Session

AlphaEvolve is an evolutionary AI system developed at Google DeepMind that automatically optimizes software solutions and discovers new algorithms. It iteratively explores variations to maximize user-defined objectives, leveraging an ensemble of language models to propose code improvements alongside evolutionary techniques. The architecture supports multi-file program optimization, allowing code to be generated and evaluated seamlessly either locally or scaled across cloud infrastructure. AlphaEvolve has already driven breakthrough real-world discoveries, including the first improvement to Strassen's matrix multiplication algorithm in 56 years, enhanced data center scheduling, and optimized hardware circuit designs. This session will provide an overview of how AlphaEvolve works, share successful use cases, and feature practical examples and live demos, and will show how you can use AlphaEvolve to optimize your scientific/technical computing applications for maximum performance and optimal results!

Speaker