Roman Korchagin

Hi! I am a Full-Stack Machine Learning Engineer specializing in building end-to-end AI-powered web applications, intelligent data pipelines, and production-ready automation.

With a strong background in Computer Science and Applied Mathematics, I bridge the gap between machine learning models and high-performance, user-friendly full-stack software.

What I can help you with:

• Machine Learning & AI Engineering:
- Custom predictive models and time-series forecasting (PyTorch LSTM, scikit-learn).
- Multimodal data processing, OCR (EasyOCR), computer vision (OpenCV), and classification pipelines.
- Seamless AI and LLM integration into your existing systems.

• Full-Stack Web Development:
- High-performance asynchronous backends with Python (FastAPI, asyncio, Pydantic, PostgreSQL/SQLite).
- Modern, responsive frontend dashboards and analytics interfaces using React 18, TypeScript, and Recharts.
- Clean REST APIs optimized for live model inference and low latency.

• Web Automation & Telegram Bots:
- Production-grade Telegram bots with aiogram 3 (complex FSM workflows, webhooks, and inline controls).
- High-throughput web parsers and scrapers (Playwright, aiohttp) built to handle anti-bot protections and rate limits.

• DevOps & Deployment:
- Full containerization with Docker and Docker Compose for easy, one-command deployment.
- Maintainable, modular code architecture with strict type safety.

Recent Project Highlights:
- Nostradamus: Full-stack demand forecasting web platform featuring automated time-series predictions (PyTorch), asynchronous FastAPI endpoints, and dynamic React 18 visualization.
- Enterprise Multimodal Sentinel: AI pipeline combining MIME parsing, OCR, and computer vision with high ROC-AUC and F1 benchmarks.

I focus on delivery speed, clean maintainable code, and proactive communication. Feel free to reach out with your project requirements — let's turn your vision into working software!