Architected a fault-tolerant, memory-to-database web scraping pipeline. The system orchestrates distributed Selenium nodes to extract unstructured automotive data and leverages LangChain and OpenAI to intelligently parse the HTML reports into actionable maintenance schedules. Engineered with robust step-level fault tolerance, ensuring seamless recovery and persistence using PostgreSQL without local file I/O.
"}},{"node":{"frontmatter":{"title":"AI Diagnostic Trouble Code (DTC) Estimator","tech":["Python","GenAI","AI Model","Data Engineering"],"github":"","external":""},"html":"Developed an asynchronous AI worker that processes vehicle engine trouble codes (DTCs). The system scrapes real-time automotive repair forums and technical service bulletins, injecting the raw unstructured data into an AI model. By engineering strict system prompts, the LLM analyzes historical repair patterns, calculates probabilities of confirmed fixes, and generates structured JSON estimating labor hours and parts costs.
"}},{"node":{"frontmatter":{"title":"Predictive 10-Year Cost Forecasting Engine","tech":["Python","Data Modeling","PostgreSQL"],"github":"","external":""},"html":"Built a purely algorithmic data modeling engine to calculate the Total Cost of Ownership (TCO) for vehicles. The engine processes immediate repairs, calculates indicator tasks, and builds a 10-year year-by-year predictive financial forecast. It factors in variable economic inflation rates, historical risk contexts (e.g., salvage records), and estimated annual mileage to generate highly granular cost breakdowns by vehicle system and priority.
"}},{"node":{"frontmatter":{"title":"Document RAG Systems","tech":["Python","GenAI","LangChain","FAISS","Ollama"],"github":"","external":""},"html":"Architected robust Retrieval-Augmented Generation (RAG) pipelines for querying unstructured PDF and TXT documents. The system seamlessly handles document ingestion, intelligent text chunking, and semantic embedding using Ollama models. Indexed via a highly efficient FAISS vector database, the pipeline leverages LangChain to orchestrate fast, context-aware question answering with local LLMs.
"}},{"node":{"frontmatter":{"title":"Stateful LangGraph ReAct Agent","tech":["Python","GenAI","LangGraph","Agentic Workflows"],"github":"","external":""},"html":"Designed a stateful, multi-step reasoning agent using LangGraph. The project breaks down complex workflows into modular nodes, allowing a local LLM to dynamically generate action plans and invoke custom Python tools. The architecture supports complex, non-linear reasoning loops and can visually map execution paths for precise debugging and scaling.
"}},{"node":{"frontmatter":{"title":"Angular CLI MCP Agent","tech":["Python","GenAI","LangGraph","Model Context Protocol (MCP)","Ollama"],"github":"","external":""},"html":"Engineered an autonomous ReAct agent using LangGraph that interfaces with a Model Context Protocol (MCP) client. The local LLM (Ollama) dynamically discovers and executes Angular CLI commands to reason about and answer complex technical prompts, demonstrating advanced tool-use and agentic orchestration over standard Input/Output streams.
"}}]}}}