📫 Email: [email protected] | OSS: [email protected]
🌍 Website: mahyar24.com
I am an experienced quantitative data scientist with seven years of industry expertise in machine learning, time-series analysis, and financial modeling. My professional focus lies at the intersection of:
- Predictive modeling and Transformers
- Natural language processing (NLP) and large language models (LLMs)
- Reinforcement learning (RL) applications in finance
- Interpretable AI
- Time-series analysis for financial and algorithmic insights
- Building AI-driven trading strategies and financial systems
I am currently seeking opportunities to begin a Ph.D. in Fall 2025. My aim is to contribute to groundbreaking advancements in machine learning, interpretable AI, and finance by blending rigorous research with practical applications.
-
Neural Networks From Scratch:
A PyTorch-like framework implemented entirely in NumPy, inspired by the book "Neural Networks From Scratch." -
Persian TTS:
A state-of-the-art Persian text-to-speech system designed and trained end-to-end with FastSpeech2. -
V2Conf:
A Python CLI tool with over 100+ monthly active users for bypassing censorship, handling a monthly traffic volume exceeding 100TB+. -
Minimal ML Implementations:
A collection of fundamental ML algorithms implemented from scratch, including GPT-2.5 (C), Logistic Regression, Decision Trees, and more. -
Persian Sentiment Analysis:
Developed a high-accuracy sentiment classifier by fine-tuning a BERT model on a large Persian text corpus, deployed in Docker as an API. -
Wabbit:
A compiled programming language created from scratch using Python and LLVMLite, inspired by David Beazley's compiler course. -
Order Management System (OMS):
Designed and contributed to a low-latency OMS for a brokerage, focusing on scalability and performance in algorithmic trading.
-
Profitable BTC Underlying Prediction from Option Data Using a Hybrid XGBoost-N-Beats Model with Venn-Abers Calibration, Conformal Prediction, and an RL Execution Agent
A novel approach to BTC price prediction, combining ML, RL, and calibrated confidence intervals for trading decisions. -
Leveraging Exchange-Rate Dynamics for Sequential Profit: A Dynamic Time Warping Portfolio Strategy in Iran’s Dual-Rate Market
Explores exploiting asynchronous market reactions in dual-rate currency regimes through dynamic time warping. -
Developing and Assessing a Long-Term Machine-Learning Trading Strategy: Transforming Time-Series Data into Classification Tabular Datasets to Simplify Data Manifold and Boost Portfolio Performance Precision with Stacking XGBoost for IREX
Introduces an innovative framework for long-term trading strategies with robust walk-forward validation.
I am an avid reader, having completed over 40 technical books on topics such as machine learning, programming, and finance. You can explore my curated list of titles here.
Feel free to reach out if you're interested in discussing research opportunities, contributing to open-source projects, or exploring new ideas in machine learning and finance. You can contact me via email at [email protected] or [email protected].