About

This is just a place for me to keep some notes and learnings on topics I'm interested in.
As I am a very visual learner, I like to create visualisations to better understand concepts.

Topics I'm currently exploring

  • Speculative Decoding - Read
  • Data Types (FP8, NVFP4, etc.) & Quantization
  • KV Cache
  • Attention Mechanisms
  • Training Techniques (SFT, RLVR, DPO, RLHF)
  • Mixture of Experts (MoE)
  • Inference Optimization (vLLM, SGLang, etc.)
  • LoRA (Low-Rank Adaptation)
  • Knowledge Distillation
  • Tokenization
  • Model Pruning

Stack / Setup

Server & Training:

  • Server: Mac Mini M1
  • DGX Spark: Training, Local Inference
  • Modal: Training, Evaluations, Sandboxes
  • Prime Intellect: Training, Evaluations, Sandboxes
Tools:
  • Harness: OpenCode
  • OpenRouter
  • AgentMail (agent communication)
  • Slack (agent communication)
  • Linear (agent communication)
  • MongoDB
  • ZVec
  • Obsidian (notes)
  • SuperMemory
  • Resend (reporting emails)