Data scientist working on statistics, machine learning, and LLMs.
I have a BSc in Computer Science from Nazarbayev University and studied Statistics and Data Science at MBZUAI. I've worked as a software engineer, data engineer, and research assistant on audio and multimodal ML. I like problems where careful statistics meets messy real-world data.
A tool-calling agent over 1,000+ Omani legal documents that answers multi-hop questions, like tracing amendment and repeal chains across laws. Combines dense + BM25 retrieval with rank fusion, cross-encoder reranking, and Neo4j graph traversal across Arabic and English.
How public attention affects calibration in prediction markets, using 679K Kalshi contracts and 72M trades. Clustered markets into five types, validated with Kruskal-Wallis and Dunn tests, and tested Granger causality against Google Trends.
Found skill-specific directions in the residual stream of Qwen3-4B, with peak separability in middle layers, and compared several steering-vector strategies on math benchmarks.
Video, audio, and text model reaching 90%+ accuracy on IEMOCAP, deployed as a real-time web app.
Python, SQL, Java, C++ · Pandas, NumPy, SciPy, scikit-learn, PyTorch · XGBoost, LightGBM, CatBoost · DuckDB · Statistical testing and experimental design