Development PracticesConference - Short25min
Hacking Performance Testing: Boilerplate as a Service
A 25-minute talk on replacing manual performance testing with meta-prompt automation. It presents a three-tier agentic loop for generating tests and automating root-cause analysis, based on scaling a metadata layer to 100k+ entities. The approach created 76 tests in minutes and cut setup time by 90%.
talk.summaryAiDisclaimer
Emilia Lendzion-BarszczSnowflake
talkDetail.whenAndWhere
Wednesday, June 17, 09:00-09:25
Room 4B
talks.roomOccupancytalks.noOccupancyInfo
Performance testing is broken. We waste 80% of our time writing boilerplate, tweaking config files (YAML or JMX), and staring at dashboards to catch a p99 spike. What if your tests could write, run, and debug themselves?
In this 25-minute talk, I will introduce "Boilerplate as a Service" - a shift from manual scripting to meta-prompt automation. Using a real-world case study of scaling a metadata layer to 100k+ entities, I’ll demonstrate how we replaced traditional CI/CD overhead with a Three-Tier Agentic Loop (Meta-Prompt Strategy, Code Generation, and Automated RCA).
You will see how this workflow generated 76 ready-to-use performance tests in minutes, cutting our setup time by 90%. We will also explore the future of debugging: moving away from generic alerts to autonomous AI agents that pinpoint the exact issue, or to heavy template rendering that kills your throughput.
Stop being a script operator. Start architecting intents.
In this 25-minute talk, I will introduce "Boilerplate as a Service" - a shift from manual scripting to meta-prompt automation. Using a real-world case study of scaling a metadata layer to 100k+ entities, I’ll demonstrate how we replaced traditional CI/CD overhead with a Three-Tier Agentic Loop (Meta-Prompt Strategy, Code Generation, and Automated RCA).
You will see how this workflow generated 76 ready-to-use performance tests in minutes, cutting our setup time by 90%. We will also explore the future of debugging: moving away from generic alerts to autonomous AI agents that pinpoint the exact issue, or to heavy template rendering that kills your throughput.
Stop being a script operator. Start architecting intents.
Emilia Lendzion-Barszcz
Emilia Lendzion-Barszcz is a Senior Software Engineer at Snowflake, specializing in high-performance backend systems and AI-driven automation. With a strong background as a QA Architect, she brings a deep Quality Mindset to development, focusing on automating workflows to compound developer efficiency rather than creating technical debt. She’s a big believer in Developer Experience (DX) and sees AI as a tool to transform engineers from script operators into intent architects. An active part of the IT community, she shares her practical experience at events like GeeCON, EuroSTAR or TestWarez, serves on Program Committees, and previously spent years co-organizing test:fest.
More on: https://www.linkedin.com/in/emilia-lendzion-barszcz
More on: https://www.linkedin.com/in/emilia-lendzion-barszcz