Yacine Mahdid Courses

Reinforcement Learning with Verifiable Rewards - RLVR Environments for LLMs

Explore Reinforcement Learning with Verifiable Rewards (RLVR) environments for LLMs, covering the complete workflow from policy rollouts to rubrics using the verifiers library.

• YouTube
• 27 minutes
• Self-Paced
• Free Video

Exploring "Logical Calculus of Nervous Activity" by McCulloch and Pitts - 1943

Dive into the foundational 1943 McCulloch-Pitts paper that predates computers yet laid groundwork for modern deep learning and neural networks.

• YouTube
• 1 hour 35 minutes
• Self-Paced
• Free Video

How to Start Learning AI Engineering Effectively as a Software Developer

Discover effective learning paths for AI engineering as a software developer, with recommended tools, projects, and resources to build your skills without analysis paralysis.

• YouTube
• 12 minutes
• Self-Paced
• Free Video

What is the Transformers' Context Window in Deep Learning and How to Make it Long

Explore transformers' context windows in deep learning, why larger windows matter, and techniques to extend them including positional encoding fixes, flash attention, sparse attention, and low-rank decomposition.

• YouTube
• 27 minutes
• Self-Paced
• Free Video

DeepSeek R1 Theory Overview - From GRPO to Reinforcement Learning and Supervised Fine-Tuning

Dive into a comprehensive breakdown of DeepSeek R1's architecture, exploring its training pipeline from GRPO and reinforcement learning to supervised fine-tuning and neural reward modeling.

• YouTube
• 26 minutes
• Self-Paced
• Free Video

KL Divergence Implementation in DeepSeek R1 - A Deep Learning Tutorial

Dive into KL divergence implementation in DeepSeek R1, exploring mathematical foundations, Monte Carlo estimation, and practical benchmarking for deep learning applications.

• YouTube
• 20 minutes
• Self-Paced
• Free Video

Group Relative Policy Optimization (GRPO) - Formula and Implementation Tutorial

Dive into Group Relative Policy Optimization (GRPO) with a detailed walkthrough of formulas and code implementation, focusing on DeepSeek R1 architecture and HuggingFace post-training techniques.

• YouTube
• 24 minutes
• Self-Paced
• Free Video

How to Serve Big LLM over Decentralized GPUs - Parallax and Dynamic Programming

Discover how to deploy large language models across decentralized GPUs using Parallax's innovative scheduling methodology and dynamic programming techniques.

• YouTube
• 37 minutes
• Self-Paced
• Free Video

Joint Self-Supervised Compression Method for ARC-AGI Problem Solving

Explore a controversial compression method achieving 27.5% on ARC-AGI v1 for just $2, featuring an in-depth interview with author Mithil Vakde on joint self-supervised learning.

• YouTube
• 1 hour 12 minutes
• Self-Paced
• Free Video

Understanding R1-Zero-Like Training with Dr. GRPO Algorithm

Explore R1-Zero-like training mysteries with Dr. GRPO algorithm's first author, covering LLM post-training, self-reflection detection, and algorithmic improvements in this deep-dive interview.

• YouTube
• 1 hour 9 minutes
• Self-Paced
• Free Video

Why Keeping Up with AI Is Exhausting - Understanding AI Misinformation and Overwhelm

Understand why AI news feels overwhelming and learn to navigate misinformation, hype cycles, and the constant stream of updates in the rapidly evolving AI landscape.

• YouTube
• 16 minutes
• Self-Paced
• Free Video

Exploring GRPO Through the RAFT Algorithm - RLHF and RLVR

Dive into GRPO's mechanics through RAFT algorithm analysis, comparing RLHF and RLVR paradigms from 2023-2025 studies to develop simplified reinforcement learning approaches for LLMs.

• YouTube
• 1 hour 4 minutes
• Self-Paced
• Free Video

How to Contribute to Open Source AI with Zero Experience

Discover how to contribute to open source AI projects with zero experience using PufferLib as an example, featuring insights from contributor DanAdvantage's journey.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

AI Path Finding - Easy Search - Raw Problem Solving

Raw problem-solving session tackling an AI path-finding challenge using the Understand, Solve, Code framework, with real-time debugging and solution implementation.

• YouTube
• 52 minutes
• Self-Paced
• Free Video

Shannon Entropy from Theory to Python

Learn to quantify information in sequences using Shannon Entropy. Explore the theory, formula, and Python implementation of this powerful information theory metric for data analysis and comparison.

• YouTube
• 23 minutes
• Self-Paced
• Free Video