I turn complex AI concepts into enterprise-ready products.
As a product-minded AI Systems Architect, I design Production RAG platforms and multi-agent workflows that solve real business problems — not demos. My work spans UGAP, UTEX, and KNIGHT: systems meant to ship, scale, and become a growth lever.
I thrive on ambiguity. If you need a visionary idea turned into a reliable, market-ready AI solution — let's talk.
| # | Project | What it is | Stack / impact |
|---|---|---|---|
| 01 | KNIGHT | Knowledge-graph framework for verifiable multi-hop MCQ evaluation · CPAL 2026 · PyPI knight-mcq |
Neo4j · RAG eval · LLMs · arXiv:2602.20135 |
| 02 | UGAP | Unified Generative AI Platform — production RAG with 7+ vector collections & live token streaming | FastAPI · LangChain · WebSockets · PostgreSQL · Redis |
| 03 | UTEX | Smart editorial AI platform — unified API gateway orchestrating multi-modal LLM workflows | 35+ endpoints · credit billing · RBAC |
| 04 | ChatDoc | Document RAG chatbot — upload docs, query with grounded answers | RAG · retrieval + generation |
| 05 | Sentiment & Emotion App | DistilBERT sentiment tooling with Streamlit / Gradio UIs | NLP · Hugging Face · FastAPI |
# professional path
2025–2026 AI Engineer & RAG Architect @ MemaranSoft
2024–2026 AI Software Developer @ SmartEra Organization
2024 GenAI Intern @ Asr Gooyesh Pardaz
2021–2023 System Analyst & Front-End Dev @ Barsa Novin Ray & RismunMemaranSoft — AI Engineer & RAG Architect · Apr 2025 – Apr 2026
- Led architecture for enterprise AI platforms during the company’s shift to AI-driven products.
- UGAP: production RAG with 7+ specialized vector collections, adaptive query processing, real-time WebSocket streaming.
- UTEX: FastAPI + LangChain backend and API gateway (35+ endpoints), multi-modal LLM workflows, credit billing, RBAC.
Python FastAPI LangChain RAG WebSockets PostgreSQL Redis
SmartEra — AI Software Developer · Dec 2024 – Feb 2026
- Neo4j knowledge-graph extraction + fact-checking DB to validate multi-agent RAG outputs.
- Dynamic prompt refiner/optimizer bridging raw LLM capability and strict business logic.
- LLM / RAG / agent systems integrated with n8n, Zapier, LangChain / LangGraph.
Neo4j LangGraph LangChain Prompt Engineering n8n
Asr Gooyesh Pardaz — GenAI Intern · Jul 2024 – Oct 2024
- RAG QA chatbots for Telegram & Bale (LangChain, LlamaIndex, FAISS).
- Multi-class sentiment package published for pip install.
- ML UIs with FastAPI, Streamlit, Gradio.
FAISS LlamaIndex FastAPI NLP
KNIGHT Framework · Co-author · Accepted at CPAL 2026
Knowledge-graph-driven generation of verifiable, multi-hop MCQs with adaptive difficulty — scalable evaluation for LLMs and RAG without repetitive manual labeling.
PyPI: knight-mcq · arXiv: 2602.20135
Education — B.S. Computer Software Engineering, Allameh Tabataba'i University (2020–2024) · GPA 3.5
Languages: English (professional) · Persian (native) · Italian (elementary)
© 2026 · erfanshm.com · Engineer · where AI meets product