Deep Dives & Systems
A look under the hood. Beyond the UI, I love exploring scalable architectures and building real-world solutions.
Live in Production
Technical Deep Dives
Trading AI (USD-M Futures Signal)
Problem
Crypto futures trading suffers from extreme volatility, high noise-to-signal ratios, and lack of visual feature explanations.
Solution
Developed an analysis-only futures prediction suite combining XGBoost classifiers and PyTorch LSTM sequence engines running under Nvidia CUDA, paired with FastAPI WebSocket streaming and transparent SHAP explanations.
AuraVision (Assistive Spatial AI)
Problem
Visually impaired individuals face dangerous physical navigation challenges and lack real-time context of spatial changes.
Solution
Built a zero-latency spatial AI system using yoloe-26N-seg object locators, multiprocessing face identification queues (GIL-free), Gemini-3-Flash multimodal scene understanding, and an adaptive RAG memory using LlamaIndex.
The Experiment Lab
Prototypes, micro-SaaS MVPs, and lessons from failed builds.