This course develops the computational thinking and algorithmic foundations required for intelligent systems. Learners study data structures, complexity, search, sorting, dynamic programming, graph algorithms, constraint satisfaction, symbolic reasoning, planning, and sequential decision-making. Classical algorithms are connected to contemporary applications such as retrieval, recommendation, constrained generation, routing, planning, and agent decision loops.

This course develops the core discipline of predictive machine learning. Learners frame business and scientific problems, prepare reliable datasets, build regression and classification models, engineer features, compare statistical and machine-learning approaches, and evaluate generalization. The course emphasizes leakage prevention, imbalance handling, calibration, uncertainty, bias-variance diagnosis, error analysis, and reliable model handoff.

This course extends machine learning into advanced representation, recommendation, temporal, probabilistic, causal, and sequential-decision systems. Learners apply clustering, dimensionality reduction, anomaly detection, recommender systems, ranking, time-series forecasting, causal reasoning, probabilistic modelling, and reinforcement-learning foundations to complex applied problems. The reinforcement-learning and decision-theoretic foundations developed here provide the basis for the alignment techniques covered in course TECH 501 Generative AI & LLM Engineering and the agentic decision-making covered in TECH 502 Agentic AI, Multi-Agent Systems and Orchestration.

This course develops learners’ ability to design, train, and evaluate advanced neural systems. It covers neural-network foundations, backpropagation, optimization, regularization, convolutional networks, sequence models, attention, transformers, representation learning, multimodal learning, graph neural networks, transfer learning, and efficient inference. Learners apply these architectures to vision, language, temporal, recommendation, multimodal, and connected data problems.

This course examines how contemporary foundation models and Generative AI systems are designed, trained, aligned, adapted, and evaluated. Learners study tokenization, embeddings, transformer internals, pre-training, scaling laws, mixture-of-experts architectures, context windows, instruction tuning, fine-tuning, PEFT, preference optimization, multimodal models, reasoning systems, and test-time compute. The emphasis is on understanding and engineering model behaviour rather than merely consuming hosted APIs.

This course develops learners’ ability to design reliable agentic and AI-native systems. It integrates agent harness design, planning, tool use, structured outputs, memory, reflection, workflow orchestration, human oversight, multi-agent collaboration, interoperability, non-deterministic testing, maintainability, and controlled autonomy. The emphasis is on framework-independent design and robust integration with enterprise systems. Retrieval and memory are addressed here as design patterns the agent reasons over; the underlying retrieval and memory infrastructure is engineered in TECH 503 AI Data Infrastructure: Pipelines, Retrieval & Knowledge Graphs.

This course develops the data infrastructure required for machine learning, deep learning, foundation models, retrieval systems, and agents. Learners design batch, distributed, streaming, feature, training data, retrieval, RAG, knowledge-graph, and memory pipelines. The course emphasizes data quality, provenance, labelling, synthetic data, contamination control, metadata, lineage, privacy, governance, and scalable knowledge access. Building on the agent memory and retrieval design patterns introduced in TECH 502 Agentic AI, Multi-Agent Systems and Orchestration, this course engineers the underlying retrieval, knowledge-graph, and memory systems that production agentic applications depend on.

This course develops the engineering capabilities required to train, deploy, scale, secure, monitor, and operate production AI systems. Learners study distributed training, GPU and accelerator systems, model serving, high-performance inference, containers, orchestration, MLOps, LLMOps, observability, reliability, security, cost engineering, capacity planning, and governance. The course treats AI as critical infrastructure rather than a standalone software feature.

This course provides significant emerging areas of AI that do not yet warrant dedicated core courses or that extend core program content into specialized domains. Because the AI field evolves rapidly, the topics offered are reviewed and refreshed regularly to reflect current research, industry, and policy developments. Learners engage with one or more thematic modules selected from a rotating set of offerings — spanning emerging technical paradigms (e.g., sovereign AI, edge AI/TinyML, causal AI, federated learning, quantum AI) and responsible AI governance (e.g., regulation, algorithmic auditing, and responsible scaling).

The capstone course requires learners to design, build, train or adapt, evaluate, deploy, scale, secure, and defend a complete AI system. Projects may involve predictive machine learning, recommender systems, deep learning, foundation models, multimodal AI, retrieval, agents, or AI infrastructure. A submission based only on calling an external API, creating prompts, or presenting a conceptual architecture is not sufficient.