Reading List¶
Priority 1: Must Read 🔴¶
Papers¶
| Paper | Year | Topic | Why |
|---|---|---|---|
| Attention Is All You Need | 2017 | Transformers | The foundation — everything builds on this |
| LoRA: Low-Rank Adaptation | 2021 | Fine-tuning | Standard method for efficient fine-tuning |
| InstructGPT (Ouyang et al.) | 2022 | RLHF | Defined the SFT → RM → PPO pipeline |
| DPO: Direct Preference Optimization | 2023 | Alignment | Simplified RLHF without reward model |
| DeepSeek-R1 | 2025 | RL/Reasoning | GRPO, on-policy RL, reasoning emergence |
Tech Blogs¶
| Blog | Topic | Why |
|---|---|---|
| Scaling LLM Post-Training at Netflix | Post-training infra | Direct insight into Netflix's stack |
| vLLM: Easy, Fast, and Cheap LLM Serving | Inference | The dominant serving engine |
Priority 2: Important 🟡¶
Papers¶
| Paper | Year | Topic |
|---|---|---|
| FlashAttention | 2022 | Efficient attention |
| QLoRA | 2023 | Quantized LoRA |
| Llama 2 | 2023 | Open-weight LLMs, RLHF details |
| Llama 3 | 2024 | Scaling, post-training at scale |
| Mixtral / MoE | 2024 | Mixture of Experts |
Tech Blogs¶
| Blog | Topic |
|---|---|
| Netflix Tech Blog (all ML posts) | Recsys, personalization, infrastructure |
| Spotify Engineering Blog | Recsys, Spark, ML platform |
| Meta AI Blog | Llama, open-source ML |
Priority 3: Deep Dives 🟢¶
Papers¶
| Paper | Year | Topic |
|---|---|---|
| FSDP (Zhao et al.) | 2023 | Distributed training |
| RoPE (Su et al.) | 2021 | Positional encoding |
| Constitutional AI | 2022 | Anthropic's alignment approach |
| Verl | 2024 | RL training framework |
Courses¶
| Course | Platform | Topic |
|---|---|---|
| Stanford CS229 (Machine Learning) | YouTube | ML foundations |
| Stanford CS224N (NLP with Deep Learning) | YouTube | Transformers, attention |
| Stanford CS336 (LLMs from Scratch) | YouTube | Building LLMs |
| Hugging Face Course | huggingface.co | Practical fine-tuning |
| Fast.ai | fast.ai | Practical deep learning |
Books¶
| Book | Author | Topic |
|---|---|---|
| Designing Machine Learning Systems | Chip Huyen | MLOps, production ML |
| Deep Learning | Goodfellow et al. | Theory foundations |
| Reinforcement Learning | Sutton & Barto | RL foundations |