Atlanta Java Users Group

​We use the Luma site to host our events. SIGN UP HERE

RAG is the pattern used to improve LLM accuracy and minimize AI hallucinations. But most implementations are a patchwork of batch embedding jobs, external vector databases, message brokers, miscellaneous services, and brittle glue code. The result is predictable: higher latency, tougher and more costly operations.
In this talk, you’ll see a streaming RAG architecture built natively in Java: continuous ingestion and transformation with distributed DAG pipelines, horizontally scalable embedding inference, and in-memory distributed vector collections for millisecond semantic search. We’ll show how data can be vectorized as it arrives from CDC, events, REST sources, and documents, how partition-aware processing cuts network overhead, and how co-locating compute with vector storage enables fast retrieval with filtering and enrichment in a single runtime. The goal is a unified, stateful platform that reduces architectural sprawl while improving latency and resilience.

Key Discussion Points:

• Why most enterprise RAG systems are not truly real-time
• Trade-offs between batch, micro-batch, and streaming AI architectures
• How streaming pipelines enable continuous ingestion and embedding
• Architectural patterns for distributed ML inferencing at scale
• How in-memory vector collections support resilient low-latency semantic search

​We use the Luma site to host our events. SIGN UP HERE

​We use the Luma site to host our events. SIGN UP HERE

Spotify's backend monorepo consolidates over 4,000 services, libraries, and components into a single repository built with Bazel, deployed to Kubernetes, and supported by 131 custom developer tools. This talk explains why Spotify chose a monorepo for its backend infrastructure, the tooling ecosystem that makes it practical — from the CLI wrapper and BuildBuddy remote execution to Declarative Infrastructure and automated monitoring — and how AI skills and agents have become essential to operating at this scale.

I will discuss the over 90 Claude Code skills that encode institutional knowledge into repeatable workflows, with a deep dive into /import-repo: a skill that has migrated thousands of polyrepos into the monorepo, evolved through 206 commits over 8 months, and taught us hard lessons about AI agent autonomy, context window limits, and the value of encoding failure patterns as guardrails.

​We use the Luma site to host our events. SIGN UP HERE

WE USE LUMA FOR EVENT SIGNUP: REGISTER HERE

# ​Knowns and Unknowns: Testing AI-Generated Code

​
​Let's face it, AI-assisted coding is rapidly becoming the primary way applications are being written. The power of writing a complex application or feature with just a few commands is too alluring to pass up.
​But between the imprecise nature of language and LLMs' ability to "hallucinate", this creates many opportunities for bugs and unintended behavior to be introduced. Just telling an AI coding agent to "write tests" isn't enough to build confidence in an application's correctness, either.
​In this presentation, we will review three distinct phases of testing AI-written code. The known knowns: using spec-driven development to define the business-meaningful behavior of your application. Known unknowns: using property-based testing to handle edge cases and avoid overfitting and unintended biases. And finally, unknown unknowns: using static analysis and mutation testing to avoid introducing bugs you didn't think about.

​AI-assisted coding is a rapidly evolving field, but this presentation should provide you with a framework and approach to build confidence when using AI to assist with writing applications.

WE USE LUMA FOR EVENT SIGNUP: REGISTER HERE

# ​DEVNEXUS April 5-7, 2027

​Devnexus is the largest and longest-running community-driven developer conference in the USA. 2027 has its strongest focus ever on helping our community prepare for, navigate, and thrive in our AI-driven dev world.

• ​Full-day training
• ​10 concurrent tracks
• ​1500+ developers
• ​Learn more at devnexus.com

WE USE LUMA FOR REGISTRATION. SIGN UP HERE

​IDEs have long provided ways to refactor code. Despite their effectiveness, that journey is arduous and time-consuming. Reluctance to refactor increases the cost of development. However, refactoring for its own sake can also lead to greater productivity losses. In this presentation, we will take a highly interactive approach. We will take an example codebase, identify modernization opportunities, use AI to refactor the code from a modernization perspective, and assess the tool's effectiveness and how we can guide it to achieve better quality and improved results.

WE USE LUMA FOR REGISTRATION. SIGN UP HERE

# ​​DEVNEXUS April 5-7, 2027

​Devnexus is the largest and longest-running community-driven developer conference in the USA. 2027 has its strongest focus ever on helping our community prepare for, navigate, and thrive in our AI-driven dev world.

• ​Full-day training
• ​10 concurrent tracks
• ​1500+ developers
• ​Learn more at devnexus.com