I'm an AI Engineer and Data Scientist with 7+ years of experience designing and deploying production-grade intelligent systems at scale. I specialize in cutting-edge AI technologies that drive real business impact:

🤖 Agentic AI & Autonomous Workflows

• Building autonomous AI agents capable of multi-step reasoning and complex task orchestration

• Designing structured reasoning frameworks for root-cause analysis and problem-solving

• Developing retrieval-augmented agentic workflows with external tool integration

💬 Large Language Models (LLMs) & NLP

• Fine-tuning GPT, LLaMA, and custom transformer models for domain-specific applications

• Implementing RAG (Retrieval-Augmented Generation) architectures with hybrid retrieval strategies

• Building Natural-Language-to-SQL agents and conversational AI systems

• Advanced prompt engineering and model optimization techniques

🔍 Enterprise RAG & Knowledge Systems

• Creating enterprise-grade knowledge engines with hybrid retrieval (semantic + keyword)

• Implementing re-ranking pipelines for improved relevance and accuracy

• Building production RAG systems for organizational knowledge management

📊 Production ML & MLOps

• End-to-end ML pipeline development from data ingestion to model deployment

• Scalable recommendation systems for large-scale e-commerce platforms

• Real-time inference systems with optimized performance and latency

Data Scientist at Daria Hamrah Paytakht (Jul 2024 – Present)

Leading AI initiatives and building production systems that serve enterprise customers:

🧠 Agentic AI Workflows

• Developed autonomous AI agents capable of performing multi-step root-cause analysis on customer complaints through structured reasoning and retrieval orchestration

• Built agents that autonomously navigate complex decision trees and generate actionable insights

💬 LLM-Powered Intelligence Systems

• Built autonomous Natural-Language-to-SQL agent capable of understanding Persian queries, generating validated SQL commands, executing on Postgres, and producing automated analysis with visualizations through an end-to-end LLM-driven workflow

• Developed LLM-powered call-center intelligence pipeline integrating speech-to-text transcription, entity extraction, and automated agent-performance scoring—substantially improving insight coverage and quality-control effectiveness

• Integrated LLM agents into analytics dashboards, enabling conversational insights, automated reporting, and interactive data exploration

🔍 Enterprise RAG Knowledge Engine

• Created enterprise RAG knowledge engine using hybrid retrieval (semantic + keyword) and advanced re-ranking to enable accurate, context-grounded responses and improved access to organizational knowledge

• Implemented retrieval pipelines optimized for accuracy and relevance in production environments

📈 Production ML Systems

• Built large-scale hybrid recommender system (content-based + collaborative) enhanced with RFM-based personalization to deliver precise and real-time user targeting

• Designed aspect-based sentiment analysis framework to surface issue-level signals across device models, directly supporting after-sales strategy and product optimization

🎤 Multimodal AI Capabilities

• Added production-grade STT and TTS capabilities for automated report narration, customer-support voice responses, and enhanced call-center automation pipelines

• 🥈 2nd Place in Tehran Provincial AI Competition (2022)

• 🎓 Member of Iran's National Elites Foundation

• 📜 Kaggle Notebooks Master

• 📄 Published Research in Health Science Reports (Wiley), ICVPR, AMLAI

OpenAI GPT LLaMA Transformers LangChain LlamaIndex RAG Fine-tuning Prompt Engineering Semantic Search Vector Databases

Agentic AI Multi-step Reasoning Retrieval Orchestration Structured Reasoning Task Orchestration Tool Integration

Transformers BERT GPT LLaMA Natural Language Processing Semantic Search Entity Extraction Sentiment Analysis Aspect-based Analysis STT/TTS

Retrieval-Augmented Generation Hybrid Retrieval Re-ranking Vector Embeddings Knowledge Graphs Document Processing

FastAPI Docker Kubernetes MLOps Model Deployment Real-time Inference A/B Testing Monitoring & Logging

Python TensorFlow Keras PyTorch scikit-learn Pandas NumPy SciPy Collaborative Filtering Deep Learning Time-Series Forecasting Causal Inference

SQL PostgreSQL NoSQL MongoDB Parquet Apache Spark Distributed Computing

React TypeScript JavaScript REST APIs GraphQL Microservices

Power BI Tableau Data Visualization Business Intelligence

End-to-end collision prediction platform using Nexar's state-of-the-art BADAS-Open model, a FastAPI backend, and a fully bilingual React dashboard.

Key Features:

• 🎯 State-of-the-Art Collision Prediction: Integrates Nexar's BADAS-Open vision model for real-time risk analysis.
• 🚀 Production-Ready Architecture: Scalable FastAPI backend and a modern, responsive React + TypeScript frontend.
• 🌐 Bilingual UI (English/Persian): Features real-time language switching and a dark, modern theme.
• 🔒 100% Offline Inference: Runs entirely locally without external API calls, ideal for production and edge deployments.
• 📊 Comprehensive Evaluation Pipeline: Includes industry-standard metrics like AUC-ROC and Average Precision.
• 🎬 Live Demo GIF: Showcases the full user workflow from video upload to risk visualization.

Tech Stack: Python | FastAPI | React | TypeScript | PyTorch | Computer Vision | MLOps

Production-Ready Features:

• Complete MLOps workflow from SOTA model integration to interactive UI.
• Designed for scalability, clean code practices, and type safety.
• Fully reproducible setup and evaluation instructions.

🔗 View Repository →

Production-ready hybrid recommender system combining collaborative filtering & content-based ML for large-scale e-commerce applications.

Key Features:

• 🔀 Hybrid recommendation engine (Collaborative Filtering + Content-Based)

• 🚀 FastAPI backend with async support and structured logging

• 🌐 Bilingual React UI (English/Persian) with RTL/LTR support

• 📊 Comprehensive offline evaluation metrics (Precision@K, Recall@K, NDCG@K, MAP@K)

• 🐳 Docker containerization for easy deployment

• 📈 Real-time recommendations with optimized sparse matrix operations

• 🔧 Modular architecture following production best practices

Tech Stack: Python | FastAPI | React | TypeScript | NumPy | SciPy | scikit-learn | Docker

Results: Achieved 140% improvement in precision and 175% improvement in recall compared to a baseline on a 38K+ user dataset—demonstrating effectiveness on challenging real-world data.

Production-Ready Features:

• Complete end-to-end pipeline from raw data to web interface

• Scalable architecture designed for enterprise deployment

• Comprehensive evaluation framework for model comparison

🔗 View Repository →

Enterprise-grade intelligent pricing and ETA prediction platform for ride-hailing platforms, combining predictive demand forecasting, real-time ETA estimation, and dynamic surge pricing optimization.

Key Features:

• 🎯 Predictive Surge Pricing Engine: Anticipates future demand-supply imbalances using ML models, enabling proactive pricing adjustments before demand spikes occur—reducing price volatility by 30-40%

• ⏱️ Advanced ETA Prediction: Achieves 20%+ improvement in accuracy over baseline using distance, speed, and zone-specific load factors with robust fallback mechanisms

• 📊 Demand Forecasting: Predicts supply-demand imbalance within 15% margin of error for 5-30 minute time horizons, enabling data-driven pricing decisions

• 💰 Revenue Optimization: Demonstrates +10-25% improvement in platform revenue per trip while maintaining customer satisfaction through balanced pricing strategies

• 🚀 Real-Time Marketplace Dashboard: Interactive React dashboard with heatmaps, KPI delta cards, and scatter plot visualizations for policy comparison and real-time monitoring

• 🔀 Multiple Pricing Policies: Implements three sophisticated pricing strategies:

  • Base Policy: Fixed pricing baseline for comparison
  • Smart Surge v1: Reactive surge pricing with demand-supply ratio analysis
  • Predictive Surge v2: Anticipatory surge pricing using short-term demand forecasting

• 📈 Policy Simulation Engine: Comprehensive replay simulator comparing pricing policies with detailed KPI analysis (ETA, completion rates, revenue, volatility)

• 🗺️ Geospatial Intelligence: Built with GeoPandas and OSMNX for Tehran, enabling accurate routing, distance calculations, and zone-based demand analysis

Tech Stack: Python | FastAPI | React | GeoPandas | OSMNX | NetworkX | scikit-learn | Time-Series Forecasting | Geospatial Analysis | MLOps

Performance Metrics:

• ✅ +20% ETA accuracy improvement over baseline
• ✅ ±15% demand forecast error margin
• ✅ +5-15% trip completion rate in high-demand zones
• ✅ +10-25% revenue efficiency per trip
• ✅ -30-40% price volatility reduction

Production-Ready Features:

• Complete end-to-end ML pipeline from data ingestion to real-time API deployment
• Scalable FastAPI backend with async support and structured logging
• Bilingual React dashboard (English/Persian) with real-time visualizations
• Comprehensive evaluation framework with policy comparison and KPI tracking
• Production-grade architecture designed for enterprise ride-hailing platforms

🔗 View Repository →

Advanced CNN-based model for highly accurate classification of blood cells. Achieved over 99% accuracy, ensuring precise identification across diverse cell types for streamlined medical diagnostics.

Tech Stack: Python | TensorFlow | Keras | CNN | Medical Imaging | Computer Vision

Production recommendation system using collaborative filtering and content-based filtering techniques. Demonstrated 8% sales increase after deployment, showcasing real business impact.

Tech Stack: Python | Collaborative Filtering | Content-Based Filtering | scikit-learn | Production ML

Automated data collection pipeline for historical stock price data from Yahoo Finance with database storage. Built for training ML models for stock price prediction and time-series analysis.

Tech Stack: Python | Web Scraping | Database Design | Data Pipeline | Time-Series Data

Deep learning model using Convolutional Neural Networks (CNNs) to classify images from the CIFAR-10 dataset. Achieved over 90% accuracy on the test set.

Tech Stack: Python | TensorFlow | Keras | CNN | Computer Vision | Image Classification

Production-ready deep learning classifier for cats and dogs using Convolutional Neural Networks. Achieved over 90% accuracy on the test set.

Tech Stack: Python | TensorFlow | Keras | CNN | Image Classification | Transfer Learning

Comprehensive customer segmentation and personality analysis for targeted marketing campaigns. Demonstrates advanced analytics and data-driven decision making.

Tech Stack: Python | Customer Analytics | Marketing Analytics | Clustering | Data Visualization | Business Intelligence

• M. Navaei, Z. Doogchi, F. Gholami, M. Kermanizadeh Tavakoli. "Leveraging Machine Learning for Pediatric Appendicitis Diagnosis: A Retrospective Study Integrating Clinical, Laboratory, and Imaging Data." Health Science Reports (Wiley)

• M. Navaei, Z. Doogchi. "Machine Learning Models for Predicting Heart Failure: Unveiling Patterns and Enhancing Precision in Cardiac Risk Assessment." Insights of Cardiovascular Pharmacology Research (ICVPR)

• M. Navaei, M. Pahlevanzadeh. "Forecasting Next-Time-Step Forex Market Stock Prices Using Neural Networks." Advances in Machine Learning & Artificial Intelligence (AMLAI)

Leading AI initiatives for enterprise customers, building production-grade systems serving thousands of users.

Led a team of 7 data scientists and analysts, overseeing end-to-end ML projects and ensuring timely delivery of scalable solutions. Implemented data-driven strategies that increased bookstore sales by 5%.

Developed ML models for book sales prediction across Iran's bookstores, enabling targeted book distribution based on reading interests in different provinces.

• Master's Degree in Artificial Intelligence | Islamic Azad University (Jun 2024 – Present)

• Bachelor's Degree in Information Technology | University of Applied Science and Technology (Feb 2024)

• 🇬🇧 English – Duolingo English Test: 120 (Proficient)

• 🇩🇪 German – A2 (Basic)

• 🇮🇷 Persian – Native

I'm actively seeking opportunities to:

• 🚀 Build and scale Agentic AI systems and LLM applications at innovative companies

• 💼 Work on production-grade AI systems that solve real business problems

• 🌍 Collaborate with international teams on cutting-edge AI/ML projects

• 📈 Contribute to enterprise RAG systems and knowledge management platforms

• 🤝 Join forward-thinking organizations that value innovation and technical excellence

Open to: Remote positions, Contract work, Full-time opportunities worldwide

I'm always open to discussing AI/ML projects, collaborating on interesting initiatives, or exploring new opportunities. Let's connect!