Agentic RAG running Mistral Small 24B with vLLM on AWS EKS, featuring LangGraph workflows, pgvector retrieval and persistent agent memory.
Fully local agentic RAG using Qwen3-8B-AWQ and vLLM on NVIDIA Jetson, orchestrated with lightweight K3s Kubernetes.
End-to-end tracing of agentic RAG workflows, providing visibility into LLM calls, tool execution and agent behaviour through Langfuse.
Automated LLM evaluation using DeepEval to assess answer relevance, tool correctness and task completion in agentic workflows.
GPU-optimised local RAG inference using NVIDIA TensorRT-LLM and Llama 3.1 on Jetson, integrated with LangGraph.
Agentic research assistant for exploring medical literature using LangGraph, OpenAI models, pgvector retrieval and persistent conversational memory.