{"data":{"_id":"6ae40d55ad69f3eadd1b3c30","name":"AI/ML","slug":"ai-ml","image":"https://images.unsplash.com/photo-1677442136019-21780ecad995?auto=format&fit=crop&w=1200&q=80","icon":"ai-ml","accentColor":"#7C3AED","isFree":false,"price":1399,"features":["PyTorch 2.x Deep Learning Core","Transformers & Self-Attention Architecture","LLM Fine-Tuning with LoRA & QLoRA","High-Throughput Inference with vLLM"],"tags":["Interactive Roadmap","AI/ML","PyTorch","Transformers","LLMs","Deep Learning"],"category":null,"description":"Master deep learning with PyTorch 2.x, build Transformers from scratch, fine-tune Large Language Models using LoRA/QLoRA, and deploy high-throughput inference servers.","modules":[{"icon":"🔥","title":"Deep Learning Foundations with PyTorch 2.x","description":"Tensors, computational graphs, Autograd automatic differentiation, custom datasets, and CUDA GPU acceleration.","duration":"3 hours","order":1,"isFree":true,"isPublished":true,"isProject":false,"content":"### The PyTorch Deep Learning Standard\nPyTorch 2.x introduces torch.compile, fusing tensor operations into fast machine code kernels for maximum GPU utilization.","subsections":[{"title":"Tensors, Autograd & GPU Acceleration","description":"Tensor operations, backpropagation through autograd, and moving models to CUDA devices.","duration":"75 mins","content":"PyTorch tensor mechanics.","isFree":true,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c79","locked":false},{"title":"Custom Datasets, DataLoaders & Training Loops","description":"Write robust training and validation loops with loss computation, optimizers, and gradient clipping.","duration":"75 mins","content":"Production PyTorch training pipeline.","isFree":true,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c7a","locked":false}],"_id":"6ae3ed503f5d6ea8c6dd3c78","locked":false},{"icon":"🧠","title":"Neural Network Architectures: CNNs to Sequence Models","description":"Convolutional layers, pooling, residual connections (ResNet), RNNs/LSTMs, and regularization techniques....","duration":"3 hours","order":2,"isFree":false,"isPublished":true,"isProject":false,"content":"","subsections":[{"title":"Convolutional Networks & Residual Connections","description":"Convolutions, feature maps, and ResNet block architectures for computer vision.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c7c","locked":true},{"title":"Sequence Modeling & Vanishing Gradients","description":"Recurrent Neural Networks, LSTMs, gating mechanisms, and sequence-to-sequence pipelines.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c7d","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c7b","locked":true},{"icon":"⚡","title":"The Transformer Architecture from Scratch","description":"Self-Attention, Scaled Dot-Product Attention, Multi-Head Attention, Positional Encodings, and Encoder-Decoder blocks....","duration":"3.5 hours","order":3,"isFree":false,"isPublished":true,"isProject":false,"content":"","subsections":[{"title":"Self-Attention & Multi-Head Attention in PyTorch","description":"Implement Query, Key, Value projections, attention masking, and scaled matrix multiplication.","duration":"90 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c7f","locked":true},{"title":"Positional Encodings & LayerNorm","description":"Sinusoidal positional embeddings, RoPE (Rotary Position Embeddings), and Pre-LN vs Post-LN.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c80","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c7e","locked":true},{"icon":"🤗","title":"The HuggingFace Ecosystem & Pre-trained Models","description":"Tokenizers (BPE, WordPiece), Transformers library, AutoModel classes, and fine-tuning on custom tasks....","duration":"3 hours","order":4,"isFree":false,"isPublished":true,"isProject":false,"content":"","subsections":[{"title":"Byte-Pair Encoding (BPE) Tokenization","description":"Vocabulary construction, special tokens, padding, truncation, and attention masks.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c82","locked":true},{"title":"Transfer Learning with HuggingFace Trainer","description":"Fine-tune BERT and RoBERTa models for sequence classification and sentiment analysis.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c83","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c81","locked":true},{"icon":"🎯","title":"Large Language Model Fine-Tuning: LoRA & QLoRA","description":"Parameter-Efficient Fine-Tuning (PEFT), Low-Rank Adaptation (LoRA), 4-bit Quantized LoRA (QLoRA), and SFTTrainer....","duration":"3.5 hours","order":5,"isFree":false,"isPublished":true,"isProject":false,"content":"","subsections":[{"title":"LoRA & QLoRA Mathematical Principles","description":"Rank matrices W = W_0 + (B * A), NF4 quantization, and double quantization.","duration":"90 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c85","locked":true},{"title":"Fine-Tuning with TRL and Unsloth","description":"Prepare instruction datasets, configure LoRA hyperparameters, and train on a single GPU.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c86","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c84","locked":true},{"icon":"🚀","title":"High-Throughput Model Serving: vLLM & Quantization","description":"PagedAttention, continuous batching, vLLM engine, model quantization formats (AWQ, GGUF), and Triton server....","duration":"3 hours","order":6,"isFree":false,"isPublished":true,"isProject":false,"content":"","subsections":[{"title":"PagedAttention & Continuous Batching with vLLM","description":"Deploy OpenAI-compatible inference servers with dynamic batch scheduling.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c88","locked":true},{"title":"Quantization Strategies: AWQ, GPTQ & GGUF","description":"Compress 70B parameter models to run on cost-effective enterprise hardware without quality degradation.","duration":"75 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c89","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c87","locked":true},{"icon":"🏆","title":"Capstone: End-to-End Specialized Domain LLM Fine-Tuning & Serving","description":"Build, fine-tune, evaluate, and deploy an enterprise specialized LLM using PyTorch, LoRA, HuggingFace TRL, and vLLM inference serving....","duration":"4.5 hours","order":7,"isFree":false,"isPublished":true,"isProject":true,"content":"","subsections":[{"title":"Instruction Dataset Engineering & QLoRA Fine-Tuning","description":"Format, tokenize, and train the model with evaluation loss tracking.","duration":"90 mins","content":"","isFree":false,"isPublished":true,"order":1,"_id":"6ae3ed503f5d6ea8c6dd3c8b","locked":true},{"title":"Model Evaluation, Quantization & vLLM Deployment","description":"Evaluate output quality, export merged weights, and serve with vLLM.","duration":"90 mins","content":"","isFree":false,"isPublished":true,"order":2,"_id":"6ae3ed503f5d6ea8c6dd3c8c","locked":true}],"_id":"6ae3ed503f5d6ea8c6dd3c8a","locked":true}],"details":[],"status":"published","__v":0},"message":"Course fetched successfully"}