AI/ML Advanced Course
AI and ML from first principles to the frontier - fully code-level, in your browser, even offline. A true beginner finishes by training their own GPT.
What it covers and how it works
A self-paced, fully code-level program that takes a true beginner from foundations to the frontier of modern AI. You build everything yourself - a micro-autograd engine, an MLP, a CNN, self-attention, and finally a working GPT trained from scratch - then learn to ship it. Eight tracks, 64 chapters, eight Forge Labs, and a portfolio-grade capstone.
Every chapter follows the same seven-step rhythm - from a real problem, to intuition, to building it from scratch, to the earned math, to the production framework, to an auto-graded challenge, to a worked failure - so knowledge compounds into capability.
A 7-step rhythm
- Why this exists - open on a real problem
- Intuition - plain language and one interactive visualization
- Build from scratch in a live code cell
- The math, now earned
- Rebuild it with the production framework
- Auto-graded challenge
- Worked failure - one realistic footgun and its fix
Learn by building
A context-aware AI mentor reads your code and hints without handing over answers. Interactive visualizations make the intuition concrete, and mastery-gated progression moves you through ranks - Apprentice, Smith, Artisan, Engineer, Architect, Master, Grandmaster - as you demonstrate real capability. It runs in your browser, even offline.
8 tracks · 64 chapters · 8 Forge Labs · 1 capstone
Each track ends in a Forge Lab that turns theory into a working artifact. Expand any track for its chapters and lab.
T1 Foundations: Code, Math & Data 8 chapters
Start here - no experience needed. Become someone who can manipulate vectors and reason about data.
- What learning-from-data means
- Python you actually need
- NumPy thinking: vectors & arrays
- The geometry of data
- Matrices as transformations
- Calculus you can see
- Probability for ML
- Loading & cleaning real data
Forge Lab: a cleaned dataset + your first hand-built classifier
T2 Classical Machine Learning 8 chapters
The workhorses, from scratch then scikit-learn. Understand bias/variance before deep learning.
- Linear regression from scratch
- Gradient descent, truly understood
- Logistic regression & classification
- Regularization
- k-NN & the curse of dimensionality
- Decision trees from scratch
- Random forests & gradient boosting
- Unsupervised: k-means & PCA
Forge Lab: a cross-validated model that beats a baseline on real data
T3 Neural Networks from Scratch 8 chapters
The conceptual heart - build a working autograd engine and train an MLP with no framework.
- The neuron
- Forward pass of an MLP
- Backprop by hand
- Build a micro-autograd engine
- Activations & initialization
- Training dynamics
- Optimizers from scratch
- Regularization for nets
Forge Lab: an MLP on your own autograd hitting strong accuracy
T4 Deep Learning with PyTorch 8 chapters
Transfer everything to the industry-standard framework. Become employable.
- From your autograd to PyTorch
- nn.Module & the training loop
- Datasets, DataLoaders, batching
- Convolutions from scratch → nn.Conv2d
- CNNs for vision
- Transfer learning
- Debugging deep nets
- Experiment discipline
Forge Lab: a fine-tuned vision model with a reproducible config
T5 Sequences, Embeddings & Attention 8 chapters
The conceptual runway to transformers.
- Representing text
- Embeddings
- RNNs from scratch
- Why RNNs struggle
- Attention from first principles
- Self-attention
- Positional encoding
- Multi-head attention
Forge Lab: a from-scratch self-attention layer with visualized maps
T6 Build a Transformer / GPT from Scratch 8 chapters
The headline outcome. A beginner who started at Track 1 builds a working GPT.
- The transformer block
- Stacking into GPT
- Tokenization for real
- Training a tiny GPT
- Sampling & generation
- Scaling laws
- Fine-tuning & instruction-tuning
- RLHF & preference tuning
Forge Lab: your own GPT generating coherent text, with a playground
T7 The Modern Frontier 8 chapters
Hands-on contact with what's current - maintained and updated regularly.
- Diffusion models from scratch
- Vision transformers & multimodal
- Retrieval-Augmented Generation (RAG)
- Agents & tool use
- Quantization & efficient inference
- Mixture-of-experts & modern architectures
- Evaluation that isn't fooled
- Interpretability
Forge Lab: a working RAG assistant over your own documents
T8 Shipping & ML Engineering 8 chapters
Turn knowledge into a career outcome.
- From notebook to service
- Data & training pipelines
- Monitoring & drift
- Cost, latency & real tradeoffs
- Responsible AI in practice
- Capstone, part 1
- Capstone, part 2
- The interview & the field
Forge Lab: a deployed, documented, portfolio-grade project
Explore the track
How you learn
- Self-paced with lifetime access
- Mentor-led cohorts
- Private corporate and enterprise delivery
- Certificate and digital badge
AI/ML Advanced Course - answered
Who is the AI/ML Advanced Course for?
True beginners who want to learn AI and ML properly, and working engineers who want first-principles depth. It starts with no experience assumed and ends with you training your own GPT and shipping a deployed project.
Do I need prior AI or math experience?
No. Track 1 builds the code, math, and data foundations from zero. The math is introduced only after you have built the thing it describes, so it is earned rather than assumed.
Is it really code-level?
Yes. Every chapter has you build the concept from scratch in a live code cell, then rebuild it the production way with the standard framework. You write a micro-autograd engine, an MLP, a CNN, self-attention, and a GPT.
What will I have built by the end?
A cleaned dataset and hand-built classifier, a cross-validated model, an MLP on your own autograd, a fine-tuned vision model, a from-scratch self-attention layer, your own GPT with a playground, a RAG assistant, and a deployed, documented capstone.
Does it run offline?
Yes. It runs in your browser, even offline, so you can learn anywhere.
Is there a certificate?
Yes. You earn a Durga Analytics certificate and a digital badge, and progress through mastery ranks as you demonstrate capability.
Is there a downloadable brochure?
Yes, from the download button at the top of this page.
Can my team take it together?
Yes. It is available as a private corporate cohort. Use the enquire button to scope it.
AI/ML Advanced Course for you or your team
Enquire about enrolment, or scope a private corporate cohort.