Ildar A. N.

Hello! My name is Ildar. I'm Data Scientist with 2+ years in commercial IT-consulting & predictive modeling + 3.5 years in R&D for energy modelling problems + MSc in Applied Math & CS.
Experinced in ad-hoc analysis, searching-for bottlenecks in modeling pipelines, statistics/ML/math model prototyping (incl. from scratch).
Proficient in Python 3 (NumPy, SciPy, Pandas, Scikit-Learn, Matplotlib, Plotly, re, Statsmodels, XGBoost, Streamlit, CloudScraper/BeautifulSoup, FiPy), PostgreSQL; Matlab/Maple, VBA. Author of three peer-reviewed publications on the mathematical modelling.

Email: abdulin.ildar.ai@gmail.com (open for new projects)    
LinkedIn: linkedin.com/in/ildar-an

📝 Main Publications

1. Production forecast method based on statistical analysis of a small sample of production data for an unconventional formation, Oil Industry Journal (Scopus), 2021.
2. Identification of fractal properties and parameters upscaling of layered heterogeneous medium, Oil Industry Journal (Scopus), 2020.
3. Numerical investigation of radial steady-state fluid flow model with Riesz potential, American Institute of Physics Conference Series (Scopus), 2020.
Full list: Google Scholar.

📈 Experience

🎓 Education and Certifications

💼 Projects — Click for Details

▶ 1. IT Consulting | Marketing Mix Modeling (MMM) Engine via Multi-Dimensional Optimization. | Python 3 • Objective: Develop an automated computational engine to quantify marketing channel decay effects (Adstock) and evaluate media investment ROI on brand search volume.
• Implementation: Engineered a custom predictive framework in Python 3 using Linear Regression combined with SciPy's Differential Evolution algorithm for global non-linear optimization. Designed a sophisticated multi-objective loss function minimizing the Euclidean norm of concatenated metric errors (R², SAE, MAPE) simultaneously across Train and Test splits (50/50) to prevent overfitting.
• Outcome: Replaced arbitrary manual parameter tuning with an automated hyperparameter search pipeline for multidimensional Adstock decay constants. Delivered a production-ready, 1-click execution pipeline with robust data preprocessing (automated schema numerical alignment, imputation, and time-series mapping) and visual diagnostic reporting. [GitHub] ▶ 2. IT Consulting | Marketing Research Data Pipeline. | Python 3 • Objective: Conduct calculations for consolidated analytics on large-scale marketing research datasets.
• Implementation: Engineered a flexible data pipeline using Python and Pandas to automate data normalization and programmatic calculation of marketing metrics.
• Outcome: Consolidated 100+ unaligned Excel files into a single unified master file, eliminating manual aggregation and standardizing data structure for downstream analysis. [GitHub] ▶ 3. IT Consulting | PoC for Real Estate Insights: Parsing and searching-for undervalued commercial properties. | Python 3 • Objective: Identify systematically undervalued commercial properties in high-potential Moscow districts
• Implementation: Built CIAN.ru web scraper (BeautifulSoup/CloudScraper) + XGBoost valuation model with expert-defined feature engineering
• Outcome: Delivered 3 expert-validated investment opportunities within MVP period, reducing noise by 40% through NLP filtering [ETL], [Modeling] ▶ 4. Gazprom Neft | Model Development for Oil Well Production Forecasting: Nonlinear Differential Equation, Numeric Methods. | Python 3 | 08/2022 – 01/2023 • Objective: Develop open-source alternative to commercial simulators for production forecasting under reservoir engineering constraints
• Implementation: Designed finite element solver (FiPy) implementing pressure-dependent permeability modeling (2D Diffusion Equation), benchmarked against tNavigator
• Outcome: Achieved 95-99% parity with industry-standard simulator results while eliminating arbitrary parameter tuning [GitHub: Solver Implementation] ▶ 5. Gazprom Neft | Excel-to-VBA Forecasting Model Transformation. | VBA | 08/2021 – 01/2022 • Objective: Modernize legacy spreadsheet-based forecasting model (600+ formulas across 7 sheets)
• Implementation: Developed automated formula parser converting Excel logic to modular VBA with unified GUI interface
• Outcome: Eliminated manual calculation errors and improved operational efficiency through single-interface solution [GitHub: Conversion Module] ▶ 6. Gazprom Neft | Ad-hoc Analysis for Small Group of Oil Wells: Regression Model Development and Monte Carlo Sampling. | Python 3 | 03/2020 – 09/2020 | [1] – Publication • Objective: Quantify key production influencers and develop probabilistic forecasting framework
• Implementation: Multivariate regression analysis + Monte Carlo simulation
• Outcome: Identified 5 statistically significant production drivers with SHAP value interpretation ▶ 7. Research lab. GAMMETT | Fractal Model Architecture for Oil Production: Inverse Problem, Testing on Real Data. | Maple | 2020 | [2, 4] – Publications • Objective: Incorporate fractal properties of the medium into a mathematical model for oil pressure.
• Implementation: Designed a new model in collaboration with a Physicist Scientist. Tested using real-world data.
• Outcome: The use of the model can improve the accuracy of predicted values by up to 20%. ▶ 8. Research lab. GAMMETT | New Model Numeric Method: Singular PDE. | Maple/Python 3 | 2019 | [3] – Publication • Objective: Develop stable numerical scheme for integro-differential equation
• Implementation: Created adaptive mesh refinement algorithm handling solution singularities
• Outcome: Published convergence proof and field validation results ▶ 9. Research lab. GAMMETT | Fractal Derivatives in Oil Production Modeling: Monte Carlo algorithm prototyping, Differential Equations. | Maple / Python 3 | 2018 | [5, 6] – Publications • Objective: Develop and compare new modeling approach (fractal derivative) for physical model with differential equations (oil production)
• Implementation: Prototyped a Monte Carlo algorithm article [arxiv.org/abs/0906.0676]. Developed a model with fractal derivatives in collaboration with two Research Scientists.
• Outcome: Conducted comparative analysis with a simpler model. [EN Description]

🛠️ Skills