Deep dives into real-time AI performance at **Automate365**, cryptographic security, and my published research on **YOLOv3**.
Lessons from building an end-to-end distributed LLM fine-tuning pipeline with QLoRA, DeepSpeed ZeRO-3, and model merging via SLERP — achieving 41.2% per-GPU memory reduction.
Implementing a robust threshold cryptography system with modular arithmetic and $2^{127}-1$ prime field security.
Upgrading a basic SSS implementation with finite field optimizations and robust error handling.
Technical breakdown of the LiveWire audio pipeline and WebSocket optimization.
How I built a production RAG system combining Qdrant dense search, Elasticsearch BM25, and Neo4j graph retrieval with a CRAG self-correction loop that improved faithfulness by 23.7%.
Engineering a Chrome MV3 extension that captures both tab audio and microphone simultaneously and streams to a Python WebSocket backend for real-time Whisper transcription.