Cohort Program

Deep Learning with PyTorch: Theory, Implementation, and Deployment

Eight weeks from first principles to fine-tuned transformers in production

About this cohort program

Deep learning courses either go too shallow, skipping the maths that makes the difference, or too deep, burying you in theory before you've shipped anything. This course gives you both: the intuition and the mathematics, tied directly to production PyTorch code you build and deploy yourself.

What you will learn

  • Build neural networks from scratch in PyTorch with no black boxes, understanding every layer and weight update
  • Implement CNNs, RNNs, attention mechanisms, and transformers with working code you can deploy immediately
  • Train models efficiently using gradient checkpointing, mixed-precision training, and distributed data loading
  • Fine-tune a pre-trained transformer on your own dataset and serve it via a production API endpoint

Who this is for

  • Software engineers and data scientists who want to understand deep learning at a fundamental level, not just use libraries
  • ML practitioners who use frameworks as black boxes and want to know what is actually happening under the hood
  • Researchers and engineers building custom model architectures who need both strong theory and deployable code

By the end

Before

Using deep learning frameworks as black boxes

After

Understanding every layer, weight update, and architectural decision

Before

Theory with no connection to working code

After

Implementations you build and deploy from the first session

Before

Prototype models that never reach production

After

Fine-tuned transformers served via monitored production APIs

Syllabus

  1. Session 1 · Saturday, August 1, 2026

    Foundations: Mathematics, PyTorch Tensors, and Autograd

  2. Session 2 · Saturday, August 8, 2026

    Building Neural Networks: Architecture, Activations, and Optimisers

  3. Session 3 · Saturday, August 15, 2026

    Regularisation, Convolutional Neural Networks, and CNN Architectures

  4. Session 4 · Saturday, August 22, 2026

    Recurrent Networks, Attention Mechanisms, and Transformers from Scratch

  5. Session 5 · Saturday, August 29, 2026

    Pre-Trained Transformers, Transfer Learning, and Fine-Tuning BERT and GPT-2

  6. Session 6 · Saturday, September 5, 2026

    Training at Scale: Mixed Precision, Gradient Checkpointing, and DDP

  7. Session 7 · Saturday, September 12, 2026

    Model Evaluation, Interpretability, and ONNX Export

  8. Session 8 · Saturday, September 19, 2026

    Serving, Monitoring, and Capstone: Fine-Tune and Deploy a Transformer

About Dr.

Dr. Kenji Nakamura

Dr. Kenji Nakamura

Deep Learning Research Scientist, NVIDIA Alumni

Vetted by Maram

Kenji spent four years on the deep learning research team at NVIDIA where he worked on training infrastructure for large-scale vision and language models. He holds a PhD in Computer Science from MIT and has published at NeurIPS, ICML, and ICLR. He now teaches deep learning at the graduate level and consults for AI labs in Tokyo and Singapore.

View full profile →

What learners say

Reviews appear here once 3 learners have completed this session.