Nebius AI Performance Engineering

This project contains my notes for a comprehensive set of materials, guides, and exercises from Nebius for AI Performance Engineering course.

I'm updating it as I go through the course.

Modules

1. From AI Model to AI Product

Explore how to transition from raw machine learning models to functional AI-driven products.

• Intro to AI & LLMs: An essential introduction to the landscape of Large Language Models. This module covers:
  • What changed with LLMs and their core limitations.
  • The GPT assistant training pipeline (Pretraining, SFT, RLHF).
  • Tokenization strategies and token economics.
  • Prompt and Context engineering techniques, including Zero-shot, Few-shot, and Chain-of-Thought prompting.
  • Practical insights into tool use and function calling.

2. LLM Architecture

Deep dive into the underlying architecture of modern LLMs.

3. MLOps

Best practices for deploying, monitoring, and maintaining machine learning models in production.

4. Performance Engineering

Techniques for optimizing model inference, reducing latency, and managing compute resources efficiently.

5. AI Model Finetuning with RL

Advanced topics in model refinement using Reinforcement Learning.