Building Production-Grade Agentic AI Systems with LangGraph
Exploring stateful multi-agent orchestration using LangGraph's interrupt/resume patterns for complex, multi-step reasoning tasks in enterprise environments.
Engineer & Maker. Exploring the intersection of Agentic AI & Java, handcrafted woodworking, South Indian cuisine, and gardening from Gilroy, CA.
12+
Years Engineering
50+
AI / Java Projects
Tech Feed
Exploring stateful multi-agent orchestration using LangGraph's interrupt/resume patterns for complex, multi-step reasoning tasks in enterprise environments.
Project Loom's virtual threads (JEP 444) enable millions of concurrent tasks without the overhead of OS threads. A deep dive into structured concurrency and ScheduledThreadPoolExecutor replacements.
Spring AI provides first-class support for vector stores, embedding models, and tool-calling. Integrating Chroma, PGVector, and OpenAI function calls in a Spring Boot application.
A framework for understanding agent architectures: tool-augmented LLMs, reflection loops, MCTS-based planning, and the role of episodic vs semantic memory in ReAct-style agents.
Compiling Spring Boot applications to native executables with GraalVM 23. Achieving cold start times under 100ms on Cloud Run with ahead-of-time compilation.
Society of Mind approaches where multiple specialized agents argue opposing positions before reaching consensus significantly reduce factual errors in complex reasoning tasks.
Building custom Kubernetes controllers using the Java Operator SDK. Reconciliation loops, status conditions, and leader election patterns for production workloads.
As models become more capable, the bottleneck shifts from prompt wording to system design: context window management, retrieval precision, and graceful degradation strategies.