About Me

Mohammad Shojaei

Machine Learning Engineer - NLP & Large Language Models

shojaei.dev@gmail.com
LinkedIn | GitHub | Hugging Face

Summary

I’m a hands-on AI engineer specializing in building practical large language model systems, with a proven track record of creating reliable AI solutions that solve real business problems. I’ve successfully developed and deployed systems that accurately process company documents, answer questions, and execute complex workflows, while efficiently training and fine-tuning models for production environments. My expertise lies in combining deep technical knowledge with practical experience to build trustworthy AI systems, focusing on proven approaches that deliver measurable business value through specialized information processing, automated workflows, and accurate task execution, all while actively contributing to the field through knowledge sharing and maintaining high standards for AI reliability and practicality.

Technical Skills

  • LLM Development & Training Transformer architectures, Attention mechanisms, LoRA fine-tuning, Quantization (GPTQ, AWQ)
  • Natural Language Processing BPE tokenization, Semantic search, Chain-of-Thought reasoning, Text-to-SQL generation
  • RAG & AI Agents Vector databases, Embedding optimization, Agent orchestration, Custom chunking strategies
  • MLOps & Deployment REST APIs, Production deployment, Monitoring, Hugging Face Hub integration
  • Model Security & Safety Prompt injection prevention, Bias mitigation, Responsible AI practices
  • Multimodal Systems Vision-Language models, Text-to-image generation, Cross-modal training
  • Optimization Techniques Distributed training, Mixed precision, KV Caching, Flash Attention
  • Development Tools Python, PyTorch, LangChain, FastAPI, Docker, Git

Professional Experience

LLM Engineer

No Limits Market Ltd | UAE, Remote | Feb 2025 – Present

  • Developed an AI-powered research system for financial analysis using custom-trained embeddings
  • Implemented automated report generation using structured LLM outputs
  • Built evaluation frameworks to ensure accuracy of AI-generated financial insights

ML Engineer

Alpha Neuroscience Company | Tehran, Iran Remote | May 2023 – Present

  • Developed signal processing algorithms for EEG data analysis using Python and specialized neuroscience libraries
  • Created an educational AI assistant for EEG technicians using RAG with custom medical knowledge bases
  • Implemented artifact detection models achieving 70% accuracy in identifying noise patterns
  • Fine-tuned a domain-specific medical LLM using LoRA for specialized neurological Q&A
  • Built and deployed a desktop application for EEG analysis using PyQt5 and scientific computing libraries

AI Engineer

Owlio | Italy, Remote | Sep 2024 – Jan 2025

  • Designed and implemented an educational AI agent using RAG and fine-tuned LLMs to convert lecture content into interactive learning materials
  • Developed custom chunking strategies for video transcripts to optimize retrieval accuracy
  • Implemented efficient token-based caching to reduce API costs by 40%
  • Built evaluation pipelines to measure RAG performance and content accuracy

Generative AI Consultant

Kerman Motor | Kerman, Iran | Dec 2022 – Jan 2023

  • Conducted technical workshops on implementing LLMs in industrial settings
  • Mentored development teams on RAG implementation and prompt engineering best practices
  • Developed proof-of-concept applications demonstrating practical AI applications

Selected Freelance Projects

  • Document Analysis System: Fine tuned multimodal VLM model for automated salary document processing and extraction
  • Property Valuation Engine: Developed RAG-based system integrating market data for real estate assistance
  • Computer Vision Pipeline: Implemented optimized object detection using Florence-2 for production deployment
  • Content Generation Framework: Created safety-aware prompt engineering system for automated social media content generation

Open Source Contributions

Github Projects

  • Cortex: Full-stack intelligent research companion using FastAPI, Milvus, and Next.js, featuring advanced RAG implementation with LangChain and Supabase integration.
  • RAG-Agent: Implementation of Retrieval Augmented Generation using LangChain and LangGraph, featuring FAISS vector store and arXiv document processing capabilities.
  • SuperAgent: AI-driven system for gathering and processing information from multiple sources, featuring advanced NLP capabilities and robust search algorithms for data analysis.
  • Open Notebook: AI-powered knowledge management system using Streamlit and LangChain for creating personalized knowledge bases from websites, PDFs, and custom inputs with GPT-powered Q&A capabilities.
  • DataSpeakGPT: Advanced text processing and OCR suite integrating GPT-3.5 Turbo for enhanced file reading (CSV, JSON, PDF) and image text extraction with multilingual support.
  • AdvancedWebScraper: Comprehensive web scraping toolkit featuring both Playwright-based advanced automation and BeautifulSoup-based general scraping with multiple output formats and data extraction methods.
  • Healthcare Assistant: Streamlit-based chat interface powered by GPT models for providing emotional support and stress analysis with actionable solutions for users.
  • GPT-Translator: Streamlit-based translation application supporting multiple providers (OpenAI, Groq) with PDF processing capabilities and progress tracking functionality.
  • Groogle: Python application combining Groq API and Google search capabilities for enhanced information retrieval and processing with concurrent web scraping functionality.
  • Youtube2Book: Extract transcripts from YouTube videos and structure them with AI using Streamlit and PySide6.

Hugging Face Models & Datasets

Language Models & Tokenizers:

Datasets:

  • Persian_QA: Synthetic question-answering dataset for Persian language model training and evaluation
  • SCED: Synthetic Contextual Enrichment Dataset - A specialized dataset designed for fine-tuning LLMs using web search and RAG, containing query-context-response pairs for enhanced language model training
  • Midjourney-Art-Prompts: Collection of curated prompts for AI art generation, facilitating creative applications

–

Education & Certifications

Bachelor of Science in Computer Engineering
University of Bam, May 2024

  • Relevant Coursework: Neural Machine Translation, Deep Learning Architectures

Certifications:

  • CS224N: NLP with Deep Learning – Stanford University (Feb 2024)
  • LangChain for LLM Application Development – DeepLearning.AI (Mar 2024)
  • Machine Learning Specialization – DeepLearning.AI (Jan 2024)
  • Python CS50P – Harvard University (Dec 2023)

Languages

  • Persian (Native)
  • English (Professional Working Proficiency)