Green AI with JobRunr: Sustainable training and inference on the JVM
summary
Small Spring Boot app collects session feedback, summarizes it with a local LLM (GPT OSS 20B via LangChain for Java), saves a Markdown summary, and uses JobRunner's carbon-aware scheduler to pick an optimal window between now and six hours. Recommendations include upgrading to Java 21+ (noticed gains up to Java 25), profiling with async-profiler, and scheduling to cut ~42.3% carbon; demo shows on-device execution and battery/energy effects.
AI-written summary · about
chapters (8)
- 0:00Green AI with JobRunr
- 1:16Green AI and scheduling
- 4:03Sustainable Java steps
- 7:23Ronald and JobRunr
- 8:31Audience analytics with carbon scheduling
- 13:14Scheduling jobs for minimal carbon emissions
- 15:24Demo and carbon results
- 16:31Q&A
related talks
25mGreen AI with JobRunr: Sustainable Training and Inference on the JVM
04h 12mJava ON 2024 Day 2
49mThe Future is 💚: A Journey through the Energy Efficient ..
32mOptimizing AI/ML workloads for sustainability
53mUpgrade Your Java with AI: From Legacy Code to High-Performance Workloads
31mCreate an eco-friendly world with Green Software Engineering
23mJava in the Green Cloud: Eco-Friendly Development with Kubernetes
28mBuilding Sustainable Software with Java