automation system using n8n in a first moment and after building the same using LangChain, and LangGraph to automatically ingest, summarize, prioritize customer support tickets and langsmith to monitor tha agents. To ensure data privacy and eliminate high API costs, I fine-tuned an 8-billion parameter open-source model (Qwen3.5 8 billions parameters at 4 bit) on Google Colab using a custom dataset of 10,000 labeled tickets. I then exported it as a .gguf file for local execution. I continuously monitored and optimize agent latency and token costs using LangSmith. Enterprise-Scale Semantic Search: I engineered an advanced RAG system to enable semantic search across client catalogs containing millions of items. I implemented the 1024-dimensional Jina v5 embedding model downloaded from huggingface and introduced an initial categorization step to generate expanded queries, significantly improving retrieval accuracy from unstructured, natural-language user inputs.…automation system using n8n in a first moment and after building the same using LangChain, and LangGraph to automatically ingest, summarize, prioritize customer support tickets and langsmith to monitor tha agents. To ensure data privacy and eliminate high API costs, I fine-tuned an 8-billion parameter open-source model (Qwen3.5 8 billions parameters at 4 bit) on Google Colab using a custom dataset of 10,000 labeled tickets. I then exported it as a .gguf file for local execution. I continuously monitored and optimize agent latency and token costs using LangSmith. Enterprise-Scale Semantic Search: I engineered an advanced RAG system to enable semantic search across client catalogs containing millions of items. I implemented the 1024-dimensional Jina v5 embedding model downloaded from huggingface and introduced an initial categorization step to generate expanded queries, significantly improving retrieval accuracy from unstructured, natural-language user inputs.WWWW…