AI & Machine Learning Notes

This section compiles theoretical references, engineering guidelines, and core cheat sheets covering Single and Multi-GPU workflows, Transformer attention mechanics, and statistical data modeling interfaces.

📖 Deep Dives

• GPU Training & Distributed Systems: Conceptual and practical deep dive into Single-GPU setups, Multi-GPU systems, Distributed Data Parallel (DDP), and High-Performance Computing (HPC) orchestration.
• Transformer Decoders & Attention (QKV): Inside attention mechanisms, key-query-value dimensions, tokenization embeddings, and structural LLM foundations.

---

Technical Paradigms Focus

I specialize in leveraging advanced deep-learning frameworks tailored specifically for spatio-temporal datasets: 1. Parameter Efficient Fine-Tuning (PEFT): Adapting massive foundational vision and remote-sensing models (e.g., Clay, Prithvi) on localized target variables with low weight budgets. 2. Scalable Pipelines: Integrating custom models on clusters running under SLURM or Dask schedulers with parallel multi-node resources.