All EXD content: interactive simulators, benchmark datasets, vLLM configs, and episode write-ups.
EXD
community
AI & ML interests
None defined yet.
Recent Activity
View all activity
Organization Card
🧠 EXD — Self-Directed PhD in AI
Engineering mastery through first principles, from the top down.
A deep dive into AI engineering — fine-tuning, architectures, inference optimization, and systems thinking. Work backwards from high-level concepts to fundamentals.
🗺️ Episodes
| # | Title | Video | Interactive | Article |
|---|---|---|---|---|
| 01 | Intro to EXD | 📺 Watch | — | — |
| 02 | Setup & First Inference | 📺 Watch | — | 📖 GitHub |
| 03 | Inference Benchmarking | 📺 Watch | 🚀 Simulator | 📖 GitHub |
| 04 | Performance Tuning | 📺 Watch | 🚀 Sim v2 | 📖 GitHub |
| 05 | Speculative Decoding | 📺 Watch | ⚡ Spec Decode | 📖 GitHub |
| 06 | Taking Stock | 📺 Watch | — | — |
| 07 | Tokenization & Embeddings | 📺 Watch | 🔤 Notebook | 📖 GitHub |
📦 Artifacts
| Type | Name | Description |
|---|---|---|
| 📊 Dataset | benchmark-results | Performance data from inference sweeps |
| ⚙️ Dataset | vllm-configs | Production vLLM configuration profiles |
| 📝 Article | GitHub episodes | Full write-ups for each episode |
| 📓 Notebook | episode-07-tokenization | Tokenization & embeddings deep-dive |
🌐 Links
📺 YouTube Channel · 💻 GitHub · 🏗️ @EXDai · 📚 Collection
🛠️ Focus Areas
- Model fine-tuning (LoRA, QLoRA, RLHF/DPO)
- Transformer architectures (attention variants, MoE)
- Inference optimization (quantization, KV cache, speculative decoding, compilation)
Work backwards. Understand everything.
models 0
None public yet