Karan shares a full LLM stack

- Karan posted a widely discussed LLM engineering stack mapping PySpark data prep, Unsloth fine-tuning, vLLM inference, FastAPI backends and LangGraph agents. - The thread also calls out observability tools like DeepEval and Langsmith and suggests knowledge-base additions such as OpenKB for retrieval layers. - It’s a practical reference for building LLM gateways and developer platforms that need data, inference, agents and traceability in one stack. (x.com)