I build
data systems
that scale.

Real-time cryptocurrency data pipeline — Binance WebSocket → Redpanda → TimescaleDB → dbt → Grafana, with Telegram alerts.

• Real-time ingestion of BTC, ETH, SOL, BNB trades via Binance WebSocket
• Kafka-compatible streaming with Redpanda, batch writes (200 records / 3s)
• TimescaleDB hypertables with auto continuous aggregates for 1m & 1h OHLCV
• 3-layer dbt models: staging, intermediate (candles), marts (volatility, volume)
• Telegram alerts for >3% price change and >3σ volume spikes
• One-command deployment with Docker Compose

Python ETL workflow for ingesting, validating, and structuring event data into analytics-ready tables.

• Designed modular ingest, transform, validate, and load stages
• Handled malformed records with structured logging and safer fallbacks
• Prepared clean output schemas for downstream analytics and reporting

Time-series forecasting project that predicts gold prices using engineered features and ensemble models.

• Engineered lag, rolling, and trend-based time-series features
• Benchmarked multiple models before selecting the strongest ensemble
• Compared predictions against actual prices with visual evaluation

Python business system with inventory, finance, and reporting workflows for small and medium enterprises.

• Designed relational data structures for inventory, finance, and staff
• Generated operational reports from aggregated transactional records
• Separated business rules from persistence logic for maintainability

Decentralized crowdfunding contract that demonstrates backend logic, state management, and blockchain fundamentals.

• Implemented deterministic contribution and refund state transitions
• Protected withdrawal logic behind campaign goal requirements
• Optimized storage layout with attention to gas costs

Collection of Kaggle notebooks exploring feature engineering, model selection, and competition-style evaluation.

• Built ensemble workflows with LightGBM, XGBoost, and CatBoost
• Used Optuna to tune models against stronger validation metrics
• Practiced rapid iteration on unfamiliar datasets and problem types