00
Orientation & Setup
Get the machine, the accounts and the habit right before anything else. Skip this and you will lose the same hours later, with worse temper.
4 topics ~13 h4 foundation · 0 core · 0 advanced
AI / ML & Data Science
From maths and programming foundations to deep learning, MLOps and a portfolio you can show.
Who it's for: Engineering students and working programmers moving into machine learning. Starts at setting up your machine and assumes no prior ML.
Sign in to keep your progress.
How much of it do you want?
Zero to advanced AI/ML/Data Science for ECE graduates and working engineers.
Narrow this list
Included in at least one reviewed higher-education syllabus.
All modules
14 modules · 182 topics · ~986 estimated learning hours. Each module builds on the ones before it — follow the "Comes after" chips to stay on solid ground.
Open
01
Mathematics for Machine Learning
You already did most of this in your ECE degree. This module reframes it around the three questions ML keeps asking: what shape is the data, which way is downhill, and how sure am I.
24 topics ~116 h8 foundation · 14 core · 2 advanced
See the 24 topics
- 1.1 Vectors, spaces and geometric intuition
- 1.2 Matrices as linear maps
- 1.3 Systems, rank, inverse, determinant
- 1.4 Eigenvalues and eigenvectors
- 1.5 SVD and low-rank approximation
- 1.6 Norms, projections and least squares
- 1.7 Derivatives, gradients, Jacobians, Hessians
- 1.8 Chain rule and computational graphs
- 1.9 Convexity and the shape of loss surfaces
- 1.10 Gradient descent and its variants
- 1.11 Constrained optimization
- 1.12 Probability foundations and Bayes
- 1.13 Random variables and distributions
- 1.14 Expectation, variance and moments
- 1.15 Joint, marginal and conditional distributions
- 1.16 Law of large numbers and the CLT
- 1.17 Maximum likelihood and MAP
- 1.18 Information theory essentials
- 1.19 Descriptive statistics and distribution shape
- 1.20 Sampling, standard error and confidence intervals
- 1.21 Hypothesis testing and p-values
- 1.22 A/B testing and experiment design
- 1.23 Correlation, causation and confounding
- 1.24 Bayesian inference in practice
Comes after00. Orientation & Setup
Open 02
Python — Basics to Advanced
Python is the lab bench for everything after this. Get fluent enough that the language stops being the obstacle.
17 topics ~78 h6 foundation · 8 core · 3 advanced6 of 17 topics in a reviewed syllabus
See the 17 topics
- 2.1 Syntax, types and control flow
- 2.2 Lists, tuples, dicts and sets
- 2.3 Functions, scope and arguments
- 2.4 Strings, formatting and regex
- 2.5 Files, paths and data I/O
- 2.6 Errors, exceptions and defensive code
- 2.7 Classes and objects
- 2.8 Inheritance, composition and polymorphism
- 2.9 Magic methods and operator overloading
- 2.10 Dataclasses, properties and clean design
- 2.11 Comprehensions, iterators and generators
- 2.12 Decorators and closures
- 2.13 Type hints, mypy and Pydantic
- 2.14 Modules, packaging and environments
- 2.15 Testing with pytest
- 2.16 Concurrency: threads, processes, asyncio
- 2.17 Profiling and making Python fast
Comes after00. Orientation & Setup
Open 03
Data Handling & Analysis
Most of a working data job is here. Models are the short part; getting trustworthy data into the right shape is the long part.
9 topics ~60 h3 foundation · 5 core · 1 advanced
See the 9 topics
- 3.1 NumPy: arrays, broadcasting, vectorization
- 3.2 Pandas: series, dataframes, indexing
- 3.3 Pandas: groupby, joins, reshaping, time series
- 3.4 SQL for data work
- 3.5 Data cleaning and leakage
- 3.6 Exploratory data analysis as a discipline
- 3.7 Visualization that communicates
- 3.8 Feature engineering
- 3.9 Bigger-than-memory data
04
Classical AI — Agents, Search & Knowledge Representation
The pre-neural artificial intelligence that still runs in production: what the field decided intelligence means, rational agents and their environments, systematic, heuristic and local search, and the symbolic and probabilistic ways of representing what a system knows.
14 topics ~63 h3 foundation · 11 core · 0 advanced
See the 14 topics
- 4.1 What artificial intelligence is, and the four ways to define it
- 4.2 How AI got here — and the two winters
- 4.3 Agents, rationality and PEAS
- 4.4 Task environments and the four agent types
- 4.5 Problem-solving agents and problem formulation
- 4.6 Judging a search strategy
- 4.7 Uninformed search
- 4.8 Informed search: greedy best-first and A*
- 4.9 Admissible and consistent heuristics
- 4.10 Local search and optimisation
- 4.11 Approaches to knowledge representation
- 4.12 Semantic networks, extended semantic networks and frames
- 4.13 Bayesian networks: representation and inference
- 4.14 Hidden Markov models
05
Classical Machine Learning
Tree ensembles still win most tabular problems in industry. Learn this properly before deep learning, not after.
22 topics ~104 h7 foundation · 15 core · 0 advanced
See the 22 topics
- 5.1 Framing an ML problem
- 5.2 Linear regression from scratch
- 5.3 Logistic regression and classification
- 5.4 Regularization
- 5.5 Bias, variance and model capacity
- 5.6 Cross-validation and honest evaluation
- 5.7 Metrics and calibration
- 5.8 k-nearest neighbours and distance methods
- 5.9 Naive Bayes
- 5.10 Support vector machines and kernels
- 5.11 Decision trees
- 5.12 Bagging and random forests
- 5.13 Gradient boosting: XGBoost, LightGBM, CatBoost
- 5.14 Clustering
- 5.15 Dimensionality reduction
- 5.16 Anomaly detection
- 5.17 Imbalanced data
- 5.18 Hyperparameter optimization
- 5.19 scikit-learn pipelines end to end
- 5.20 Model interpretability
- 5.21 Time-series forecasting
- 5.22 Recommender systems
Comes after03. Data Handling & Analysis
Open 06
Deep Learning
Where your signals-and-systems instincts pay off. Build the intuition and the training craft, not just the API calls.
21 topics ~117 h1 foundation · 15 core · 5 advanced
See the 21 topics
- 6.1 From perceptron to multilayer network
- 6.2 Backpropagation by hand
- 6.3 Activations and initialization
- 6.4 Loss functions
- 6.5 Optimizers and learning-rate schedules
- 6.6 Regularization and augmentation
- 6.7 Normalization layers
- 6.8 PyTorch: tensors, autograd, modules
- 6.9 The training loop, datasets, dataloaders
- 6.10 Debugging deep learning
- 6.11 Image processing fundamentals for machine learning
- 6.12 Convolutional networks
- 6.13 Modern vision architectures
- 6.14 Transfer learning and fine-tuning
- 6.15 Object detection and segmentation
- 6.16 Recurrent networks and their limits
- 6.17 Attention from first principles
- 6.18 Autoencoders and representation learning
- 6.19 Generative models: GANs and diffusion
- 6.20 Self-supervised and contrastive learning
- 6.21 Scaling training: precision, accumulation, distribution
Comes after05. Classical Machine Learning
Open 07
Natural Language Processing
The bridge between classical ML and the LLM era. Do not skip the pre-transformer half — it is where the evaluation instincts come from.
11 topics ~55 h2 foundation · 9 core · 0 advanced
See the 11 topics
- 7.1 Text preprocessing and tokenization
- 7.2 Bag of words, TF-IDF and n-grams
- 7.3 Word embeddings
- 7.4 Sequence labelling
- 7.5 Text classification end to end
- 7.6 The Transformer architecture in full
- 7.7 Encoder models: BERT and family
- 7.8 Decoder models: the GPT family
- 7.9 Seq2seq tasks
- 7.10 Hugging Face ecosystem
- 7.11 Evaluating NLP systems
Comes after06. Deep Learning
Open 08
Large Language Models
The commercially hot layer. The skill that separates people here is evaluation and cost control, not prompting.
14 topics ~81 h0 foundation · 8 core · 6 advanced
See the 14 topics
- 8.1 How LLMs are trained
- 8.2 Tokenizers, context and positional schemes
- 8.3 Decoding and sampling
- 8.4 Prompt engineering that survives users
- 8.5 Structured output and tool schemas
- 8.6 Embeddings and vector databases
- 8.7 Retrieval-augmented generation
- 8.8 Advanced RAG
- 8.9 Fine-tuning: LoRA, QLoRA, PEFT
- 8.10 Quantization, distillation and local models
- 8.11 Serving LLMs: latency, throughput, cost
- 8.12 Evaluating LLM systems
- 8.13 Safety, hallucination and guardrails
- 8.14 Multimodal models
Comes after07. Natural Language Processing
Open 09
Agentic AI
The newest layer and the least settled. Learn the patterns and the failure modes; frameworks here have a shelf life of months.
11 topics ~53 h0 foundation · 5 core · 6 advanced
See the 11 topics
- 9.1 What an agent is — and when you don't need one
- 9.2 Agent loops: ReAct, planning, reflection
- 9.3 Tools, function calling and action spaces
- 9.4 Memory
- 9.5 Model Context Protocol and interoperability
- 9.6 Agent frameworks
- 9.7 Multi-agent systems
- 9.8 Computer-use and browser agents
- 9.9 Evaluating and observing agents
- 9.10 Agent security
- 9.11 Cost control and production failure modes
Comes after08. Large Language Models
Open 10
Production & MLOps
The difference between a notebook and a product. Employers pay for this part, and most self-taught candidates skip it.
11 topics ~55 h0 foundation · 7 core · 4 advanced
See the 11 topics
- 10.1 From notebook to package
- 10.2 Experiment tracking
- 10.3 Data and model versioning
- 10.4 Serving models
- 10.5 Docker and reproducible environments
- 10.6 CI/CD for machine learning
- 10.7 Cloud fundamentals and cost
- 10.8 Monitoring, drift and retraining
- 10.9 Data pipelines and feature stores
- 10.10 Scaling inference
- 10.11 Responsible AI and Indian data law
Comes after05. Classical Machine Learning
Open 11
The ECE Bridge — Signals, Edge & Embedded AI
Your unfair advantage. Most CS graduates cannot do this half, and hardware-adjacent AI roles pay for it.
8 topics ~42 h0 foundation · 5 core · 3 advanced
See the 8 topics
Comes after06. Deep Learning
Open 12
Frontier Topics
What changed recently and is likely to still matter. Treat this module as perishable — revisit it every quarter.
8 topics ~29 h0 foundation · 1 core · 7 advanced
See the 8 topics
Comes after08. Large Language Models
Open 13
Capstones, Portfolio & Interviews
Nothing here is optional. Four finished projects beat twelve finished courses in every hiring conversation.
8 topics ~120 h0 foundation · 6 core · 2 advanced
See the 8 topics
Comes after05. Classical Machine Learning
Open