AI SOLUTIONS ARCHITECTURE

Set up your learning environment, explore the AI landscape, and get a clear roadmap for the journey ahead. You'll understand how the course is structured, what you'll build, and how each module connects to real-world AI implementation.

• Meet your instructor
• Course structure
• Assignments & final project overview
• AI landscape
• Environment set up

Cut through the hype and build a practical understanding of today's AI ecosystem. Learn why this wave of AI is different, where the opportunities are emerging, and how organisations are positioning themselves to compete.

• The evolution of modern AI
• Understanding the AI ecosystem
• Models, platforms, applications, infrastructure
• Separating hype from reality
• Emerging opportunities & risks
• Hands-on activity & quiz

Move beyond experimentation and learn how organisations successfully adopt Generative AI at scale. Discover how to identify high-value opportunities, avoid common pitfalls, and create momentum for AI initiatives.

• The enterprise AI adoption journey
• Identifying high-impact use cases
• Jobs-to-be-Done framework for AI
• Prompt-thinking fundamentals
• Common AI implementation failures
• Assessing organisational readiness

Assignment #1: GenAI Solution Proposal & Data Strategy

Go beyond buzzwords and understand what powers modern AI systems. Build the technical intuition needed to evaluate models, design AI solutions, and make informed engineering decisions.

• How Large Language Models work
• Transformers, tokens, attention
• Training & fine-tuning fundamentals
• RLHF & model behavior
• Performance, cost, latency trade-offs
• Workshop: ML model implementation

Learn why great AI starts with great data. Build the foundations needed to support scalable AI initiatives, from governance and quality to embeddings and modern data architectures.

• Why data quality determines AI quality
• Data taxonomy: Structured, unstructured, synthetic
• Data pipelines for AI: Ingestion, transformation, curation
• Embeddings & vector representations
• Building a data strategy that supports AI initiatives
• Data governance basics: Lineage, quality, access

Assignment #2: RAG System Prototype

Turn AI ambition into a clear execution plan. Learn how to prioritise initiatives, align stakeholders, and build a roadmap that balances quick wins with long-term transformation.

• The AI roadmap framework: Value, feasibility, readiness axes
• Quick wins vs. transformational bets
• Stakeholder alignment & building the business case
• AI maturity models: Where are you today?
• Sequencing initiatives across a 30/60/90 & 12-month horizon
• Metrics that matter: Measuring AI success

Master the end-to-end process of building and operating AI systems. Understand how AI delivery differs from traditional software development and where governance, evaluation, and operations fit in.

• How AI-DLC differs from traditional SDLC
• Stages: Discover → Design → Develop → Evaluate → Deploy → Monitor
• Responsible AI checkpoints at each stage
• Prompt engineering & iteration as a development discipline
• Model evaluation frameworks: RAGAS, human eval, benchmarks
• MLOps & LLMOps: Continuous delivery for AI systems

Treat prompting as an engineering discipline, not guesswork. Learn proven frameworks that improve reliability, reasoning, and performance across modern AI applications.

• Anatomy of an effective prompt
• Core techniques: Zero-shot, few-shot, chain-of-thought, role prompting
• System prompts & persona design
• Advanced patterns: ReAct, self-consistency, tree of thought
• Prompt versioning, testing, management
• Hands-on: Prompt patterns that actually work in production

Learn how to connect AI systems to trusted knowledge sources. Design retrieval pipelines that produce more accurate, relevant, and trustworthy outputs at scale.

• Why RAG exists: The knowledge cutoff & hallucination problem
• The RAG pipeline: Chunk → embed → index → retrieve → generate
• Naive vs. Advanced vs. Agentic RAG
• Chunking strategies, embedding models, vector stores
• Re-ranking, hybrid search, contextual retrieval
• Evaluating RAG quality: Faithfulness, relevance, recall

Discover when prompts and retrieval are no longer enough. Learn how to customise models efficiently to improve performance, reduce costs, and solve domain-specific problems.

• When to fine-tune vs. prompt engineer vs. RAG
• Fine-tuning methods: Full fine-tune, LoRA, QLoRA, instruction tuning
• Dataset preparation & quality criteria
• Model distillation: Teaching small models from large ones
• PEFT & parameter-efficient approaches
• Evaluating & deploying custom models

Assignment #3: Deployment Plan & Cost Analysis

Navigate the AWS AI ecosystem with confidence. Learn how to select the right services, build scalable architectures, and move AI workloads into production.

• Amazon Bedrock: Foundation model access, APIs, capabilities
• Knowledge bases, guardrails, model evaluation on Bedrock
• Amazon SageMaker: Training, hosting, pipelines
• AWS AI services: Rekognition, Textract, Transcribe, Comprehend
• Choosing the right AWS service for your use case
• Reference architectures for common AI workloads

Assignment #4: Monitoring & MLOps Strategy

Explore the shift from AI assistants to AI actors. Learn how autonomous systems reason, use tools, and execute tasks while understanding when agentic approaches truly add value.

• What makes a system "agentic"?
• Agent anatomy: LLM brain, tools, memory, environment
• Core agentic patterns: ReAct, Plan-and-Execute, Reflection, MRKL
• Single-agent vs. multi-agent systems
• When to use agents vs. simpler AI approaches
• The agentic maturity ladder

Learn how teams of AI agents collaborate to solve complex problems. Design orchestrated systems that are scalable, resilient, and production-ready.

• Multi-agent architectures: Supervisor, peer-to-peer, hierarchical
• Agent roles: Orchestrator, subagent, critic, executor
• Communication protocols: A2A, MCP, tool calling
• Handoff patterns & state management across agents
• Amazon Bedrock agents & multi-agent collaboration
• Designing for reliability: Retries, fallbacks, human-in-the-loop

Give agents the ability to remember, adapt, and improve over time. Learn the architectural patterns that transform short-lived interactions into persistent intelligence.

• Types of agent memory: In-context, episodic, semantic, procedural
• Short-term vs. long-term memory architectures
• State management: Checkpointing & persistence
• Tool use & external system integration
• Building agents that improve over time
• Practical patterns: Session memory, user profiles, task resumption

Assignment #5: End-to-End GenAI Solution

Secure AI systems before they become business-critical. Understand the threat landscape and learn how to build safeguards that protect data, models, and users.

• The AI threat landscape: Prompt injection, data poisoning, model theft
• Jailbreaks, adversarial inputs, indirect injection attacks
• Securing the AI pipeline: Input validation, output filtering, sandboxing
• Secrets, PII, sensitive data handling in LLM contexts
• AWS Bedrock Guardrails & security controls
• Zero-trust principles for agentic systems

Learn how to measure, monitor, and improve AI systems in production. Build observability practices that turn AI from a black box into an accountable business capability.

• Why AI observability is different from traditional APM
• LLM-specific metrics: Hallucination rate, faithfulness, groundedness
• Evaluation strategies: Offline evals, online A/B, human feedback loops
• Observability tools: Langfuse, Phoenix, AWS CloudWatch for AI
• Building a feedback loop from production into the AI-DLC
• Alerting, drift detection, model degradation signals

Finish the course with a leadership framework for scaling AI responsibly. Learn how to govern AI, manage risk, control costs, and prepare for the next wave of intelligent systems.

• The governance imperative: Regulation, risk, trust
• Responsible AI pillars: Fairness, explainability, accountability, privacy
• AI policy frameworks: EU AI Act, NIST AI RMF, ISO 42001
• Building an AI Center of Excellence (CoE)
• FinOps for AI: Managing token costs, compute, and ROI
• Emerging frontier & building your personal AI learning strategy

Final Project: End-to-End Agentic AI Solution