← Back to the catalogue

Level 3 · Ph.D

Advanced AI

Ph.D syllabus · 63 phases · +1 year

Every unit this programme teaches, in the order it is taught. Headlines only — the material itself opens once you are enrolled.

Dates, price and prerequisites

1. Phase 0 From builder to architect
2. Phase 1 Transformer & attention (architect-depth recap)
3. Phase 2 Modern LLM landscape
4. Phase 3 Decoding & sampling
5. Phase 4 Tokenization & context windows
6. Phase 5 Embeddings & vector spaces
7. Phase 6 Evaluation foundations
8. Phase 7 Prompt anatomy & instruction design
9. Phase 8 Reasoning I: chain-of-thought, decomposition, least-to-most
10. Phase 9 Reasoning II: self-consistency, reflection & debate
11. Phase 10 Tool-augmented reasoning: ReAct & plan-execute
12. Phase 11 Structured & constrained outputs
13. Phase 12 Hallucination I: taxonomy, root causes, when models fabricate
14. Phase 13 Hallucination II: mitigation
15. Phase 14 Prompt optimization
16. Phase 15 Prompt security
17. Phase 16 Context engineering
18. Phase 17 RAG fundamentals
19. Phase 18 Advanced retrieval
20. Phase 19 Advanced RAG: multi-hop, graph RAG, agentic retrieval
21. Phase 20 RAG evaluation
22. Phase 21 Memory systems
23. Phase 22 Feature context & context-as-a-product
24. Phase 23 Knowledge management: corpora, freshness, provenance, governance
25. Phase 24 Tool design: interfaces, affordances, error handling, safety
26. Phase 25 The agent loop: plan–act–observe; single-agent patterns
27. Phase 26 Specialized agents & roles; agent design patterns
28. Phase 27 Multi-agent orchestration I: supervisor/worker, hand-offs
29. Phase 28 Multi-agent orchestration II: parallel fan-out, deterministic workflows, pipelines
30. Phase 29 Skills & modularized AI: reusable capabilities, composition
31. Phase 30 Agent memory, state & long-running/durable tasks
32. Phase 31 Agent reliability: guardrails, validation, human-in-the-loop, recovery
33. Phase 32 Agent evaluation, observability, tracing & debugging
34. Phase 33 Architecture principles: modularization & separation of concerns
35. Phase 34 Blueprints: designing & documenting AI systems for big/complex projects
36. Phase 35 Decomposition & context boundaries; interfaces between AI modules
37. Phase 36 Architectural patterns: router, pipeline, cascade, ensemble, blackboard, fallback
38. Phase 37 Reliability & resilience: retries, circuit breakers, graceful degradation
39. Phase 38 Cost/latency/quality optimization: model routing, caching, distillation choices
40. Phase 39 Security & privacy architecture: PII, secrets, tenant isolation, system-level injection defense
41. Phase 40 Scaling complex AI projects: organizing many agents, skills & contexts; system governance
42. Phase 41 Data strategy for applied AI; pipelines; synthetic data
43. Phase 42 Fine-tune vs RAG vs prompt: decision framework; PEFT/LoRA
44. Phase 43 Evaluation at scale: offline/online, LLM-as-judge, golden sets, regression suites
45. Phase 44 LLMOps I: versioning (prompts/models), registries, CI/CD for AI
46. Phase 45 LLMOps II: monitoring, drift, observability & incident response
47. Phase 46 Productionizing: serving, scaling, cost governance & SLAs
48. Phase 47 AI strategy & value: use-case discovery, ROI, feasibility, build vs buy
49. Phase 48 Applied AI in IT/software: codegen, devops, support, internal knowledge
50. Phase 49 Applied AI across industries I: finance, healthcare, legal (constraints + compliance)
51. Phase 50 Applied AI across industries II: manufacturing, retail, education, public sector
52. Phase 51 Responsible AI & governance: risk, model cards, regulation
53. Phase 52 The AI consultant's craft: discovery, requirements, stakeholders, deliverables, roadmaps, change management
54. Phase 53 Capstone I: discovery → architecture blueprint → modular multi-agent design → eval harness
55. Phase 54 Capstone II: build & deliver — production hardening + consulting deliverable package
56. Phase 55 Frontier & what's next: agentic systems, multimodal, on-device, research signposts; portal integration & program close
57. Phase 56 Serving & deploying inference: batching, caching, autoscaling
58. Phase 57 SLOs, SLIs & error budgets for a probabilistic system
59. Phase 58 Observability & tracing for LLM & agent systems
60. Phase 59 Reliability patterns around model calls: retries, breakers, fallbacks
61. Phase 60 Eval-regression & release gating: CI/CD for models & prompts
62. Phase 61 Drift & quality monitoring + RAG pipeline ops
63. Phase 62 LLMOps/AgentOps & responsible-AI ops: the strand capstone