Bringing GenAI to the Modern Enterprise. A production use-case. In Serverless Java!
summary
A production Java RAG app is built, tested and deployed to Cloud Run using SpringAI, with a LangChain4J counterpart, storing embeddings in AlloyDB (PGVector) and calling Vertex AI’s Gemini Pro plus local models via Ollama/Testcontainers and GKE/vLLM. Features include multimodal ingestion (image OCR → embedding → function call for bookstore availability), function-calling plumbing, and observability (Prometheus/Grafana). Native images cut cold-starts from ~3.4s (JIT) to ~0.6s; summarization patterns run ~8s (prompt-stuffing), ~19s (map-reduce) and ~33s (refine). The demo processes Gutenberg books end-to-end.
AI-written summary · about
chapters (16)
- 0:00Introduction to GenAI in Enterprise
- 1:08Business-driven GenAI adoption
- 7:28Production use case overview
- 11:18GenAI fundamentals and concepts
- 25:53Architecture and technology stack
- 36:20Why Java for GenAI applications
- 41:15Testing GenAI applications
- 47:07Optimizing GenAI applications
- 1:00:51Q&A session part 1
- 1:17:26Advanced GenAI patterns and best practices
- 1:33:11Model hosting and deployment options
- 1:50:04Summarization patterns comparison
- 2:06:18Data management and vector databases
- 2:10:17Observability in GenAI applications
- 2:21:16Fun exercise and closing remarks
- 2:34:43Q&A session part 2
related talks
47mBlazingly Fast GenAI App Development Java and Spring AI
26mJava’s GenAI Era: Getting Spring AI and Neo4j Applications to Production
07mServerless: Accelerate your AI
46mGenerative AI and Java
48mFamiliar Tools in Unknown Territory: GenAI with SpringAI & Testcontainers
38mBuild Gen AI applications with Google Databases and Cloud Runtimes
46mBuilding Powerful GenAI Apps with Pure Java
52mIntelligent Beans with Spring AI