Erfan Shafiee Moghaddam.

• ·Engineered a knowledge graph extraction pipeline using Neo4j and built a dedicated fact-checking database to validate multi-agent RAG outputs and mitigate LLM hallucinations.
• ·Designed and deployed a dynamic prompt refiner and optimizer system for large-scale organizational projects, bridging raw LLM capabilities and strict business logic.
• ·Built cutting-edge AI solutions with LLMs, RAG systems, and intelligent agents; integrated models with automation platforms (n8n/Zapier) and LangChain/LangGraph workflows.
• ·Mentored teams during a startup weekend and deployed AI-powered process automation for enterprise clients.

Freelance · Tehran Province, Iran · Hybrid

Neo4j · Knowledge Graphs · LangGraph · LangChain · RAG · LLMs · Prompt Engineering · n8n · Zapier · Fact-Checking

• ·Developed advanced RAG-based QA chatbots for messaging platforms (Telegram, Bale) using LangChain, LlamaIndex, and FAISS.
• ·Engineered a multi-class sentiment analysis package and distributed it as a pip-installable Python package to enhance conversational AI.
• ·Built interactive ML interfaces with FastAPI, Streamlit, and Gradio; integrated robust APIs for real-time data exchange.

Full-time · Tehran Province, Iran · On-site

Shell Scripting · Gradio · Streamlit · FastAPI · LangChain · LlamaIndex · Retrieval-Augmented Generation (RAG) · Vector Databases · FAISS · NLP · Sentiment Analysis · Telegram API · GitHub · Python

• ·Translated ambiguous business requirements into functional software using BPMS platforms and custom C#/SQL/JavaScript logic.
• ·Developed tailored solutions on the System Builder low-code platform, including BPMS workflows and advanced SQL reporting.
• ·Designed responsive web interfaces with HTML, CSS, JavaScript, and Bootstrap—establishing a strong commitment to end-user experience.

Part-time · Tehran Province, Iran · Hybrid

C# .NET · JavaScript · SQL · BPMS · Low-Code Development · Requirements Analysis · Responsive Web Design · Bootstrap

• ·Knowledge-graph-driven framework for verifiable, multi-hop multiple-choice questions with adaptive difficulty — scalable evaluation for LLMs and RAG systems.
• ·Engineered the core Python package and open-source repo; published on PyPI (knight-mcq). arXiv: 2602.20135

Knowledge Graphs · RAG Evaluation · LLMs · NLP · Python