Working independently on AI-powered systems and open source tooling. Core focus areas:

• ScamIntelli: Designing and building a real-time fraud detection platform using graph ML, GNNs, and a FastAPI + Redis + PostgreSQL backend stack.
• GeneralsAI: Researching LLM-based game AI agents with hierarchical planning and chain-of-thought reasoning.
• WZML-X: Maintaining and extending an open source Telegram automation framework (1000+ GitHub stars) with async task queuing and a plugin-based architecture.
• Consulting: Backend architecture reviews, ML pipeline design, and infrastructure automation for small teams.

Active contributor to open source projects in the Python, automation, and bot development ecosystem.

• Architected and shipped the WZML-X async task queue, replacing a synchronous execution model and enabling concurrent multi-user operations.
• Designed the plugin system that decoupled download sources from core bot logic — enabling community contributions without requiring core code changes.
• Implemented the Docker-based deployment model and Pydantic configuration layer, reducing setup friction from hours to minutes.
• Triaged issues, reviewed pull requests, and mentored new contributors in the community.
• Contributed patches to upstream dependencies in the Python Telegram bot ecosystem.

Intensive period of building and studying across the ML and distributed systems stack.

• Completed deep study of graph neural networks, culminating in the ScamIntelli GNN detection engine design.
• Built multiple small-scale distributed systems projects to develop fluency with message queues, worker pools, and fault-tolerant design patterns.
• Studied and implemented papers on fraud detection: graph-based approaches, anomaly detection, and adversarial robustness.
• Developed Python proficiency in async/await patterns, type-annotated codebases, and production-grade error handling.
• Built and deployed personal tooling that became the foundation for WZML-X.