This project focuses on building an end-to-end time series forecasting system to predict daily grocery sales for different product families across multiple stores. The goal is to leverage historical sales data along with external factors such as promotions, holidays, and oil prices to accurately forecast future demand.

Pipeline-engineering harness for VCE Mathematical Methods 3-4: SymPy verification + bilingual lesson PDFs + cheat sheets + mock SAC papers, all reproducible from one CLI.

Automated domain-investing pipeline · AI Council valuation (Claude + DeepSeek) · hard budget walls · 22-case end-to-end smoke · MIT

ExtendScript-based Photoshop automation tool that programmatically creates structured artboards with metadata-driven naming and layout logic.

Metadata-first creative asset pipeline demonstrating workflow automation, schema governance, and production efficiency optimization.

Расчет шпунтового ограждения по ВСН 136-78 с автоматическим подбором глубины заглубления, эпюрами активного/пассивного давления, изгибающего момента и формированием инженерного отчета.

Experimented with 16+ deep neural network configurations in TensorFlow on e-commerce demand data to compare tuning sensitivity, performance, and runtime tradeoffs. Best model: 2-layer (128->64), ReLU, Adam, LR 0.001, batch 128, early stopping, validation MAE 12.54 in ~126 seconds.