Ali Afsah-Noudeh

Senior AI Engineer · AI Agents, RAG & LLM Systems

I build production AI agent systems — DAG-based orchestration, multi-agent architectures, RAG, MCP servers, and the safety controls that keep LLMs reliable when it matters: structured output validation, prompt-injection defence, hallucination mitigation, graceful degradation. 14+ years across the full stack back that up. This portfolio is a playable homage to PlayStation's Twisted Metal: you drive a white Peykan around a PS1-style arena and blow up billboards to discover my projects.

Zurvan
ACTIVE
An open-source, goal-driven Python framework for building production AI agents.
- Python
- GAME-loop architecture
- Pydantic
- LiteLLM
FULL DOSSIER
Iran Wildfire Detection
ACTIVE
Production-grade real-time wildfire detection for Iran — a fault-tolerant multi-stage agent pipeline.
- Python
- asyncio
- DAG executor
- NASA FIRMS
FULL DOSSIER
Shahnameh AI Agent
ACTIVE
A RAG agent over the 60,000-couplet Shahnameh — citation-grounded answers where frontier LLMs guess.
- Python
- Whisper large-v3
- Postgres/pgvector
- MCP
FULL DOSSIER
Book Summarizer AI Agent
ACTIVE
An agentic PDF-to-summary system with the safety controls of a production service.
- Python
- Multi-agent pipeline
- Prompt-injection defence
- LLM-as-judge
FULL DOSSIER
Personal Librarian
ACTIVE
A multilingual RAG assistant that answers questions across your whole book and document library.
- Python
- ChromaDB
- Hybrid retrieval
- Query expansion
FULL DOSSIER