Automating Code Performance Optimization with AI with Saurabh Misra

AI-assisted coding helps developers write code quickly, but ensuring that all new code and software systems are highly performant remains challenging. Writing fast code requires expertise, and optimizing performance demands significant manual effort. Consequently, much modern software runs slower than expected, leading to dissatisfied users and high cloud costs.

In this session, we'll explore how AI-based optimization, combined with automated correctness testing and benchmarking, can streamline the process of writing performant code. Through live code examples, we'll demonstrate how Python programmers can inadvertently write slow code and how they can automatically optimize both existing and future code. We'll dive into the principles that make AI-based performance optimization possible. Additionally, we'll see how it's now feasible to incorporate AI tools that continuously optimize new code, seamlessly ensuring our software remains fast by default. This allows developers to focus on writing new features while AI takes care of software performance.

1 - Code we write today can be slow without our awareness.
2 - Understanding the principles that enable AI to optimize both new and existing code for performance.
3 - Developers can now integrate AI-based automated performance optimization as a software engineering best practice.

Saurabh is the CEO and Founder of Codeflash, an AI tool that automatically finds the most performant version of your Python code through benchmarking—while verifying its correctness. He studied Machine Learning at CMU and Electronics Engineering at IIT, built performance tools at NVIDIA, and led ML products at Cresta.

Follow Codeflash at -

Optimization Pull requests referenced in the presentation -