Select Page

Generative AI presents new challenges in understanding how to test the many layers for performance. It requires training on new technologies, and a focus on building a performance engineering strategy around AI technology. Scott Moore and Daniel Geater (Qualitest) discuss the current state of AI performance engineering, and where we are headed in the near future.

The challenge for performance engineers is to understand how each component that make up AI affect the overall performance of the solution as a whole, and how that affects the end user experience.

Video Insights Into AI Performance Engineering

πŸ€– AI-specific offerings from AWS and Azure, along with new technologies like tensor flow and PyTorch Lightning, are shaping the future of performance engineering.

πŸ“Š The foundational models like GPT are trained using a huge amount of carefully curated data from massive sources, with billions of parameters and hundreds of millions of inputs.

πŸ“Š Python’s popularity in machine learning and data science is rooted in its origins in the academic world of science, math, and statistics, leading to the development of a rich ecosystem of libraries and frameworks.

πŸš€ Python’s success is due to its ease of learning and the wealth of libraries and frameworks available, making it easy to build something quickly.

πŸ“ˆ We need to focus a lot more on our performance test inputs with AI, our focus isn’t just about does a technical function scale horizontally as it’s called, but about how specific data sets provided to the same function stress its ability to process

πŸ“Š Engineers need to focus on understanding the statistical side of how AI models work and process information in order to effectively test AI performance.

🧠 The evolution of AI tooling has been spurred by advancements across academia and industry, everything from GPT models to advanced searches and social media filters, leading to more advanced searches and tooling.

🌍 Qualitest has been working with AI for 6-7 years, focusing on boosting accuracy and stability for various industries across the globe.

πŸ’™ Tricentis β–Ί Make sure to visit them and tell them β€œThank You” for making this show possible.

Want to support PERFTOUR? Buy Me A Coffee!

Connect with me


πŸ”— Links: