Master neural networks in PyTorch, from backpropagation and CNNs to Transformers, GANs and diffusion, and ship optimised, deployed deep learning models with AI-reviewed projects
A 12-week, project-based program that goes deep on neural networks themselves. After setting up Git and GitHub, you revise the maths behind deep learning, master PyTorch tensors and autograd, and build backpropagation from scratch before training networks with modern optimisers, normalisation and regularisation. You then build CNNs from LeNet to ResNet, fine-tune pretrained models, detect objects with YOLO and segment images, model sequences with LSTMs, and build attention, Transformers and a mini GPT from scratch. You fine-tune BERT, T5 and LoRA models with Hugging Face and create images with VAEs, GANs and diffusion models. Finally you make models fast and production-ready with mixed precision, distributed training, quantisation, ONNX export, FastAPI serving and experiment tracking. Every week ends with a project reviewed by AI, and you finish with a deployed capstone and interview preparation.
Category: AI & Machine Learning
Level: Intermediate
Length: 12 weeks, online and self-paced
Price: ₹5,999
Rating: 8.5/10
What you will learn
Use Git and GitHub with branches and pull requests for deep learning projects
Apply the linear algebra and calculus behind neural networks and verify gradients in code
Work fluently with PyTorch tensors, autograd and GPUs on Colab and Kaggle
Implement backpropagation and a micrograd-style autograd engine from scratch
Train deep networks with modern optimisers, learning-rate schedules, initialisation and normalisation
Fight overfitting with dropout, weight decay and augmentation, and debug networks that will not learn
Build CNNs from LeNet to ResNet and interpret them with Grad-CAM
Fine-tune pretrained models from torchvision and timm, detect objects with YOLO and segment images
Model sequences and time series with RNNs, LSTMs, GRUs and sequence-to-sequence models
Build self-attention, Transformers, a BPE tokenizer and a mini GPT from scratch
Fine-tune BERT, T5 and other Hugging Face models, including efficient fine-tuning with LoRA
Create images with autoencoders, VAEs, GANs and diffusion models
Speed up training with mixed precision, torch.compile and distributed data parallel
Compress models with quantisation, pruning and distillation and export them to ONNX and TorchScript
Serve models with FastAPI and Docker and track experiments with Weights & Biases and MLflow
Skills
PyTorch
Neural network fundamentals and backpropagation
Training and debugging deep networks
Convolutional neural networks
Transfer learning with torchvision and timm
Object detection with YOLO and image segmentation
RNNs, LSTMs and sequence models
Attention and Transformer architectures
Hugging Face Transformers and fine-tuning
Parameter-efficient fine-tuning with LoRA
GANs, VAEs and diffusion models
Mixed precision and distributed training
Model compression and ONNX export
Model serving with FastAPI and Docker
Experiment tracking with Weights & Biases and MLflow
Frequently asked questions
Do I need prior experience?
You should know Python and basic machine learning concepts. Comfort with school-level linear algebra helps.
How long is the course and how much time do I need each week?
The course runs for 3 months, with 12 weekly modules. Most learners spend 8–10 hours a week. It is self-paced, so you can go faster or slower.
How are my assignments reviewed?
You push your code to GitHub and AI reviewers check it for bugs, quality and security within minutes, with clear fixes.
Will I get a certificate?
Yes. Finish the weekly assignments and the final project to download a verified certificate with a unique ID that employers can check.
Can I ask questions if I get stuck?
Yes. Use Ask your mentor inside any week to send a question to your mentor.
What roles does this course prepare me for?
Deep Learning with PyTorch prepares you for roles such as Deep Learning Engineer, Computer Vision Engineer, Machine Learning Engineer and NLP Engineer. The projects you push to GitHub become a portfolio you can show in interviews.