Target Roles¶
Role → Skills Mapping¶
Netflix Research Scientist 5/6 (\(466K–\)750K)¶
| Requirement | Wiki Section |
|---|---|
| LLM development, post-training, fine-tuning, distillation | 1. Post-Training |
| Distributed training | 2. Distributed Training |
| Python, PyTorch, TensorFlow | Throughout |
| Spark, Hive, Hadoop | 4. Data at Scale |
| Reinforcement learning | 7. RL for LLMs |
| Personalization, recommender systems | 5. Recommender Systems |
| Production-ready systems | 8. MLOps |
| Conversational agents, search, NLP | 6. Transformers |
Other Target Roles¶
| Company | Role | Comp Range | Key Differentiator |
|---|---|---|---|
| Anthropic | Research Engineer | \(300K–\)500K | RLHF, safety, constitutional AI |
| OpenAI | Research Scientist | \(300K–\)600K | Post-training, reasoning, evals |
| Google DeepMind | Research Scientist | \(250K–\)500K | Gemini, multimodal, RL |
| Meta FAIR | Research Engineer | \(250K–\)450K | Llama, open-source, systems |
| Spotify | ML Engineer | \(200K–\)350K | Personalization, recsys, Spark |
| Apple | ML Engineer | \(250K–\)400K | On-device ML, efficiency |
Study Priority by Role Type¶
"I want to work on LLM post-training" (Netflix, Anthropic, OpenAI)¶
- Post-Training (SFT, DPO, GRPO) ⭐⭐⭐
- RL for LLMs ⭐⭐⭐
- Distributed Training ⭐⭐⭐
- Transformers ⭐⭐
- Inference & Serving ⭐⭐
"I want to work on recommendations/personalization" (Netflix, Spotify)¶
- Recommender Systems ⭐⭐⭐
- Data at Scale (Spark/Hive) ⭐⭐⭐
- Transformers ⭐⭐
- Post-Training ⭐⭐
- MLOps ⭐⭐
"I want to work on ML infrastructure" (Netflix, Google, Meta)¶
- Distributed Training ⭐⭐⭐
- Inference & Serving ⭐⭐⭐
- MLOps ⭐⭐⭐
- Data at Scale ⭐⭐
- Post-Training ⭐⭐