← Back to the catalogueDeep Learning
Classroom . Online . Hybrid
Mode
Course overview
A deep learning course covering network architectures from dense layers to transformers. Learners build networks by hand before using frameworks, so training behaviour — initialisation, activation, gradient flow, regularisation is understood rather than assumed. Vision, sequence and attention architectures each get a full module with a training project.
All work runs on TensorFlow, Keras and PyTorch with GPU sessions on Colab.
Who it is for
- Learners who have completed a machine learning course
- AI and data science professionals
- Researchers and postgraduate students
- Engineers working on vision or speech
Prerequisites
Machine learning fundamentals, Python, NumPy and basic linear algebra.
How it runs
Learn → Practise → Build → Experience → Demonstrate, ending in a capstone. Delivered by practitioners from the engineering bench, in Madurai, Coimbatore and online.
Final project
Vision or Sequence Model, Trained and Served
Programme sheet
The printed sheet carries the full module breakdown, labs, project work and certification path. Fees, dates and formats for the next intake are confirmed by the education team on enquiry.
Learning outcomes
01
Explain forward and backward passes through a network.
02
Choose activations, losses and optimisers for a task.
03
Diagnose underfitting, overfitting and vanishing gradients.
04
Build convolutional networks for image tasks.
05
Build recurrent and LSTM networks for sequences.
06
Implement attention and read a transformer diagram.
07
Train efficiently with augmentation, callbacks and transfer learning.
Module structure
8 modulesModule 01
Neural Network Basics
Perceptron . Multilayer networks . Activation functions . Loss functions . Gradient descent . Backpropagation
Lab
Network From Scratch in NumPy
Module 02
Training Deep Networks
Initialisation . Learning rate schedules . Dropout . Optimisers . Batch normalisation . Early stopping
TensorFlow . Keras sequential and functional API . PyTorch tensors . Autograd . Datasets and loaders . GPU usage
Lab
Same Model in Keras and PyTorch
Module 04
Convolutional Networks
Filters and feature maps . Data augmentation . Transfer learning
Recurrent networks . LSTM . GRU . Sequence-to-sequence . Time-series windows
Lab
Time-Series Forecasting
Module 06
Attention and Transformers
Attention mechanism . Self-attention . Positional encoding . Encoder and decoder . Pretrained transformers
Lab
Fine-Tune a Small Transformer
Module 07
Generative Models
Autoencoders . Variational autoencoders . GAN basics . Diffusion overview . Evaluating generated output
Module 08
Projects and Deployment
Experiment tracking . Model export . ONNX . Serving . Inference optimisation
Assessment & certification
Module assignments and labs
25%
Final project and review
40%
Learners who complete all modules, submit the final project and clear the review receive a course completion certificate from Kaizen Infinities Private Limited. Project work is documented for the learner's portfolio, and interview preparation is included in the closing sessions.
Career outcomes . Roles this programme prepares for
Deep Learning EngineerAI Research EngineerComputer Vision EngineerNLP EngineerMachine Learning Engineer
