Build Portfolio-Worthy Capstone Projects
Each course ends with a hands-on capstone — a real brief you build end to end. These are the actual Mastery & Capstone challenges from our AI & ML tracks, not sample titles.
7 capstone projects
- Capstone Python
Multi-Agent Workflow
AI Agents & Agentic Workflows
Complex tasks benefit from specialised agents working together. Build a two-agent pipeline where a Researcher gathers facts and a Writer formats the report.
Start the courseDeliverables
- • Implement ResearcherAgent.research(topic) returning a list of bullet-point facts
- • Implement WriterAgent.write(topic, facts) returning a formatted markdown report
- • Orchestrate both agents and print the final report
- Capstone Python
Inference Pipeline Simulator
Hugging Face Transformers
Simulate the full inference pipeline of a HuggingFace text-classification model: tokenise → pad/truncate → mock encode → decode label.
Start the courseDeliverables
- • Implement tokenise_and_pad(text, max_len=16) returning padded token IDs
- • Implement mock_classifier(ids) that returns a label based on keyword signals
- • Run the pipeline on 4 sample sentences and print results
- Capstone Python
LLM Chain Simulator
Large Language Models
LLM chains pass the output of one step as input to the next. Build a synchronous chain executor with context accumulation.
Start the courseDeliverables
- • Implement Chain.add_step(name, transform_fn) to register pipeline stages
- • Implement Chain.run(input) passing context through all stages
- • Build a 3-step summarise → translate → format chain using mock transforms
- Capstone Python
Iris Flower Classifier
Python for AI & ML
Build a complete ML pipeline: load the Iris dataset, split it, train a classifier, and evaluate it — end-to-end in under 20 lines.
Start the courseDeliverables
- • Load the Iris dataset and split into 80/20 train/test (random_state=42)
- • Train a LogisticRegression classifier
- • Print accuracy, and the classification report
- Capstone Python
CNN Forward Pass
PyTorch Deep Learning
Build a small Convolutional Neural Network and trace a batch through it — confirming each layer's output shape.
Start the courseDeliverables
- • Define SimpleCNN with Conv2d(1,16,3) → ReLU → MaxPool2d(2) → Linear(→10)
- • Run a forward pass with a batch of 4 grayscale 28×28 images
- • Print each intermediate tensor shape
- Capstone Python
End-to-End RAG Pipeline
Retrieval-Augmented Generation
Wire together all the components: chunk a knowledge base, embed the chunks, retrieve the top-k, and assemble a grounded prompt.
Start the courseDeliverables
- • Chunk the knowledge base (3 sentences per chunk, 1 sentence overlap)
- • Embed all chunks with mock embeddings and retrieve top-3 for a query
- • Assemble the final RAG prompt with [CONTEXT] and [QUESTION] tags
- Capstone Python
Full Training Pipeline
TensorFlow & Keras
Train a complete classification pipeline on synthetic data — preprocessing, model, training with callbacks, and evaluation.
Start the courseDeliverables
- • Standardise X with (X - mean) / std
- • Train the model with early stopping
- • Print accuracy on the held-out test set