Taha Bakhtari

AI Engineer & Researcher

Specializing in AI Agents, Large Language Models, and Computer Vision

Explore AI Agents View Projects
Taha Bakhtari

About Me

Born May 9, 2009, in Tehran, I am a programmer, researcher, and AI specialist focusing on cutting-edge technologies. With over two years of self-taught experience in artificial intelligence, I develop innovative solutions in areas such as Large Language Models (LLMs), Generative AI, Computer Vision, and Multi-Agent Systems.

Life & Education

I developed a passion for programming and AI from an early age. As a 10th-grade student, I've independently mastered advanced concepts in deep learning, natural language processing, and computer vision, demonstrating that age is not a barrier to innovation in technology. My work spans multiple domains, from sophisticated AI systems to practical applications that solve real-world problems.

20+ AI Projects
2+ Years Experience
4 Research Areas

AI Agents & Systems

I specialize in building advanced AI agent systems that can perceive, reason, and act autonomously in various environments.

FarsiRAG Retrieval System

An advanced question-answering system for Persian PDFs using Retrieval-Augmented Generation techniques. This system efficiently processes and retrieves relevant information from Persian documents.

View on GitHub

Advanced RAG Techniques

Implementation of cutting-edge Retrieval-Augmented Generation methods for more accurate and contextually-aware AI responses, showcasing the latest innovations in NLP technology.

View on GitHub

Subtitle Generator

An AI-powered system that automatically generates and synchronizes subtitles for videos with high accuracy. This popular project has received significant community recognition.

View on GitHub

AI CAPTCHA Solver

A deep learning system using YOLO architecture that achieves over 95% accuracy in solving CAPTCHAs, trained on a dataset of 181,000 images I personally collected and labeled.

View on GitHub

MCP Servers

Implementation of Multi-agent Chat Protocol servers for orchestrating communication between multiple AI agents, enabling complex collaborative behaviors and distributed problem-solving.

View on GitHub

TorobjoMCP

A specialized MCP implementation for product search and Instagram analysis, enabling intelligent data retrieval and social media content analysis with a custom protocol.

View on GitHub

Professional Projects

With extensive hands-on experience in AI, I have developed a wide range of advanced projects across multiple domains:

Computer Vision Projects

  • Persian digits OCR with custom datasets
  • Iranian license plate recognition system
  • Cancer detection using Convolutional Neural Networks
  • CAPTCHA solver with 95%+ accuracy using YOLO architecture
  • Custom OCR for handwritten document processing
  • Image classification for medical diagnostics
  • Object tracking in video streams

NLP & Language Models

  • Retrieval-Augmented Generation for Persian documents
  • Speech-to-text conversion for multiple languages
  • Sentiment analysis for social media content
  • Automatic subtitle generation and synchronization
  • Question-answering systems from specialized documents
  • LLM fine-tuning for specific domains
  • Text classification and categorization systems

Data Science & Predictive Analytics

  • Loan repayment prediction using financial data
  • Diabetes diagnosis from medical parameters
  • Advanced audio processing for pattern recognition
  • Large dataset processing and management
  • Custom OCR dataset creation (118k+ labeled images)
  • Predictive maintenance for equipment
  • Data visualization and analytics dashboards

AI Agents & Infrastructure

  • Multi-Agent Chat Protocol (MCP) servers
  • Autonomous decision-making systems
  • Agentic AI for task automation
  • LLM integration with external tools
  • RAG systems for knowledge-intensive tasks
  • AI agent orchestration frameworks
  • Interactive agent-environment systems

Skills & Expertise

Programming

Proficient in Python and specialized AI libraries

Deep Learning

TensorFlow, PyTorch, Neural Networks, CNN, YOLO

AI Agents

Multi-Agent systems, Agentic AI, Autonomous systems

Computer Vision

Object detection, OCR, image processing, CAPTCHA solving

NLP & LLMs

Retrieval-Augmented Generation, Language models, Transformers

Data Science

Dataset creation, data processing, predictive modeling

Model Optimization

Model quantization, pruning, and efficient deployment

AI System Design

End-to-end AI system architecture and implementation

Certificates

My certifications and achievements in AI and programming

AIOcup Certificate

2nd Place Nationwide - AIOcup

From: Aiolearn

National Creativity and Innovation Competition (ICUP) in programming and AI - Challenge A5

LLM Evaluations Certificate

Large Language Model Evaluations

From: EVIDENTLY AI

LLM evaluations for AI product teams

Hugging Face Agents Certificate

Foundations of AI Agents

From: Hugging Face

Unit 1: Foundations of Agents in the Hugging Face Agents Course

HackerRank Certificate

Problem Solving (Intermediate)

From: HackerRank

Problem Solving (Intermediate)

ID: 738EB5A5C8DE

Content Creation

I produce cutting-edge and highly practical content focused on real-world AI implementations. My goal is to teach advanced and often overlooked topics in the Persian-speaking AI community through educational and experimental videos.

YouTube Channel

Tutorials and demos on Large Language Models (LLMs), connecting LLMs to Blender and the web, building multi-modal AI agents, exploring complex AI infrastructures, and running creative AI experiments.

@Tahabakhtariai

Telegram Channel

Sharing AI news, tips, and educational content. Get the latest updates on my projects and discoveries in the world of AI technology.

@Ai_by_Taha

GitHub

Explore my open-source AI projects. All code repositories are available for you to study, use, and contribute to the advancement of AI technology.

@Tahabakhtari

Get In Touch

Interested in collaborating on AI projects or have questions about my work? Feel free to reach out through any of my social platforms or email me directly.