A highly scalable evaluation system designed to benchmark and stress-test over 100+ AI models concurrently.

• Engineered the core logic using React 18 and integrated directly with HuggingFace APIs.
• Implemented the Model Context Protocol (MCP) to provide autonomous agents with context-aware tool access.
• Deployed serverless orchestration layers via Cloudflare Workers for zero-latency cross-region requests.

A deep reinforcement learning infrastructure engineered for autonomous, dynamic volatility spot trading.

• Architected advanced Actor-Critic (PPO) models using PyTorch to navigate non-stationary market conditions.
• Developed sub-millisecond execution algorithms and high-frequency data ingestion pipelines.
• Achieved a verified +74.24% ROI during rigorous backtesting and live simulation phases.

An immersive 3D WebGL data visualization platform rendering large-scale geographic and industrial data.

• Rendered and processed 35,000+ global power plants seamlessly in-browser utilizing GeoJSON data structures.
• Optimized heavy rendering pipelines using Three.js and React Three Fiber instancing, achieving stable 60 FPS.

A complex web-based multiplayer ecosystem built for high concurrency and immediate state synchronization.

• Architected real-time client-server communication using Firebase and WebSockets.
• Developed custom rendering engines leveraging HTML5 Canvas and Framer Motion for fluid UI interactions.

An experimental R&D initiative simulating an entire company run by interconnected AI agents.

• Utilized CrewAI and Vector Databases to manage memory, role delegation, and cross-agent communication.