Stanislav M.

I develop Python applications and machine learning workflows—from preparing raw data and training models to building APIs, dashboards, and desktop tools. My work includes financial analytics software, demand forecasting, neural network classification, and a custom YOLO image segmentation proof of concept. What I can help you with: • Machine Learning: classification, regression, baseline forecasting, feature engineering, and model evaluation. • Neural Networks: training and comparing models, handling imbalanced datasets, and experimenting with regularization. • Computer Vision Prototypes: image annotation workflows, dataset preparation, and custom YOLO segmentation experiments. • Python Development: data processing tools, API integrations, analytical dashboards, and desktop applications. • Existing Projects: debugging Python code, improving notebooks, and turning experiments into reusable scripts. Selected projects: — MBS Analytics Tool: a Python desktop and CLI application for processing Bloomberg-style Excel/CSV exports, analyzing mortgage-backed securities, and generating charts and Excel reports. — DemandPilot: an inventory planning portfolio application with demand forecasting, holdout-based model comparison, stock-risk analysis, scenario simulation, and a FastAPI backend. — Neural Network Classification: TensorFlow/Keras experiments on balanced and imbalanced tabular datasets, including preprocessing, class weighting, early stopping, and evaluation with precision, recall, F1, and ROC-AUC. — YOLO Segmentation Proof of Concept: a custom image dataset annotated in CVAT, converted from COCO to YOLO format, and used for local training and visual evaluation. Tools I work with: Python, pandas, NumPy, scikit-learn, TensorFlow/Keras, PyTorch, Ultralytics YOLO, matplotlib, seaborn, FastAPI, Streamlit, SQL, Git, and Docker. I focus on reproducible workflows, clear evaluation, and practical deliverables. Depending on the project, you receive source code, setup instructions, documented results, and a usable notebook, script, API, or application. Send me your requirements, dataset, or existing code, and we can define a clear first milestone.