Week 3: Standing on the Shoulders of Giants & Building RAG

Stop reinventing the wheel. In the real world, you stand on the shoulders of giants. We explore Transfer Learning so you can hijack state-of-the-art models and fine-tune them for custom datasets. Then, we dive into Retrieval-Augmented Generation (RAG) to give LLMs a custom memory so they can chat with your private data. Finally, we master Convolutional Neural Networks and the art of "babysitting" the learning process with hyperparameter tuning. Core Topics: Transfer Learning & Fine-Tuning, Retrieval-Augmented Generation (RAG) Architecture, Convolutional Neural Networks (Kernels, Stride, Padding), Max Pooling, Mean Squared Error vs. Cross-Entropy, and Hyperparameter Optimization (Random Search, Grokking, and Overfitting). Week 3 Lab: Code a full-stack RAG search engine for 11,000 SIGGRAPH papers using FastAPI and Qdrant Cloud. Build hybrid retrieval and cross-encoder reranking to stop reinventing the wheel.

8 weeks · 59 lectures · free to watch