About Me
I am a data scientist who loves turning raw data into AI systems that deliver tangible business value. I have built fraud models, scheduled pipelines, and simple APIs. My stack includes Python, Airflow, Docker, and AWS to make workflows repeatable. My professional motto is to keep learning, build systems that act, learn, and automate useful work.
Experience
baly.iq (Rocket Internet)
Sep 2024 – Sep 2025: Data Scientist
• Deployed Docker-based services to deliver business features quickly, coordinating with tech and
ops.
• Automated fraud-data pipelines with Airflow, reducing manual work by 30%.
• Tuned databases using Percona PMM, improving query response by 20% and ensuring 99.8% HA.
• Deployed containerized services using Docker and Ansible to deliver key business features faster.
snapp.ir (Rocket Internet)
Sep 2021 – Sep 2024: Fraud Data Analyst
• Investigated 35M+ weekly transactions to detect unusual patterns and report findings.
• Built SQL detection rules and Python scripts to flag abnormal transactions, reducing processing time
by 40%.
• Collaborated with cross-functional teams to refine business rules.
• Learned automation and containerization to build scalable fraud-detection tools used in production.
Projects
Smart Telegram Bot using n8n
Built an automation workflow connecting a Telegram bot to OpenAI GPT via n8n; used ngrok for SSL and short-term context memory in chat responses.
LLM Fraud Detection
Designed a transformer-based LLM model for Arabic fraud-text detection with reproducible pipelines. [GitHub]
Interactive Model Interpretation Dashboard
Developed SHAP and feature-importance visuals with FastAPI backend and Docker deployment. [GitHub]
S3-MinIO Starter Kit
Created an S3-compatible local storage using Docker + MinIO for testing pipelines; later deployed on AWS EC2. [GitHub]
Algorithmic Trading System
Worked in a small research group building an algorithmic-trading framework for XAUUSD and major forex pairs. Processed 5-minute timeframe market data with Polars for feature generation based on price-action patterns and technical indicators. Used MetaTrader 5 in Docker containers for live data collection and feed simulation, linking it to a local database for real-time updates. Leveraged TradingView for pattern classification and Forex Factory for macro-event features such as Federal Funds Rate changes. Ran experiments and model backtests on Kaggle with the Weights & Biases tracking tool to study models’ behavior and refine their parameters.
COVID-19 Prediction
Built a predictive model for COVID-19 death counts using time series methods. [Colab]
Skills
Programming:Python, R, C, Golang, Rust (intermediate)
ML/AI Tools:TensorFlow, PyTorch Geometric, GNN, FastAPI
Databases:PostgreSQL, MySQL, Neo4j, ClickHouse
Automation:Apache Airflow, Ansible (intermediate)
Containerization:Docker
Version Control:Git
Web Technologies:CSS (Bootstrap), Flask
Soft Skills:Team leadership, structured problem-solving, clear communication, cross-team coordination.
Education
2025–2026: Master of Science, Data ScienceUniversity of Amsterdam (UvA). Coursework: Machine Learning & Optimization, Advanced Analytics for a Better World, Impact Evaluation.
2017–2019: Master of Business Administration (MBA)University of Tehran. GPA: 3.11/4. Courses in Strategy, Marketing, and Financial Management.
2012–2016: BSc, Industrial EngineeringAmir Kabir University of Technology (Tehran Polytechnic). GPA: 2.9/4. Courses in Probability, Operations Research, and Linear Algebra.
Certifications
Divide and Conquer, Sorting and Searching, and Randomized Algorithms — Coursera
Applied Statistical Modeling for Data Analysis in R — Udemy
Associate DevOps Engineer — ArvanCloud
Languages
| Language | Level |
|---|---|
| English | Fluent |
| Dutch | Basic |
| German (Deutsche Sprache) | Basic |
| Persian | Native |
Interests
Deep learning, LLM, MLOps, algorithmic trading, and Agentic AI