Prompt Engineering Mastery II: Advanced Reasoning, Agents and Prompt Optimisation

Course Prerequisite(s)

Please note that this course has the following prerequisites which must be completed before it can be accessed
Prompt Engineering Mastery: From Foundations to Future

About Course

Prompt Engineering Mastery II: Advanced Reasoning, Agents and Prompt Optimisation

You know how to write an effective prompt. Now learn how to design the programme around it: how it reasons, searches, uses examples, calls tools, improves from feedback and produces work across different media.

This free advanced prompt engineering course from NORAI Academy is the continuation of Prompt Engineering Mastery: From Foundations to Future, not a replacement for it. Complete the original course first. Named refresher references connect familiar concepts to the new mechanisms without repeating the full introductory lessons.

Go beyond a single prompt

Across 80 text lessons in 16 sections, you will study advanced reasoning and test-time compute, demonstration engineering, self-consistency, best-of-N, Tree of Thoughts, adaptive workflows and automatic prompt optimisation. Worked traces explain DSPy, MIPROv2, SIMBA and GEPA in plain English.

You will also examine tool-using agents, multi-agent cooperation, MCP and A2A concepts, evidence-led research, multimodal investigation, creative production, prompt caching, model transfer and deployment economics. Research clinics connect established methods and recent papers to decisions you can actually analyse.

Advanced depth, clear explanations

Follow Atlas, a fictional research-and-communication workbench, through contrasting designs. Inspect the inputs, intermediate records, observations, costs and checks. Then transfer those ideas to unfamiliar final-assessment organisations. Every assessed task includes its necessary facts; synthetic case results are teaching fixtures, not live model benchmarks.

Build a private mastery workbench

Sixteen worked studios and embedded worksheets help you develop task contracts, instruction stacks, demonstration policies, reasoning ledgers, search controllers, optimisation plans, agent policies, evidence matrices and a complete architecture dossier. Private practice is unsubmitted and includes worked comparisons.

Automatically marked assessments

Sixteen section checkpoints and a separate final challenge contain 192 fixed-key questions. Analyse traces, calculate stipulated costs, compare designs and choose supported conclusions. No essays, uploaded projects, instructor marking or AI grading service are required.

Pass each section at 8 of 10 and the final at 26 of 32. Attempts are unlimited and there is no time limit. Worked feedback explains every option. No paid AI subscription, live API calls or coding are needed to complete the course.

A connected three-course pathway

Start with Prompt Engineering Mastery: From Foundations to Future. Continue here with advanced prompt programmes. Context Engineering: Beyond Prompt Engineering is the next learning step for information selection, memory, state and the context lifecycle across whole workflows.

Passing the supplied analytical cases demonstrates performance on those cases. It does not independently certify production deployment competence, professional licensing or university credit.

Turn a task into a prompt programme with clear contracts, observations and acceptance conditions.
Engineer demonstration sets and allocate reasoning effort under a declared budget.
Trace self-consistency, best-of-N, Tree of Thoughts and adaptive controllers.
Understand automatic prompt optimisation, DSPy, MIPROv2, SIMBA and GEPA through worked examples.
Design tool-using agents, specialist handoffs and bounded multi-agent cooperation.
Develop evidence-led research, multimodal investigation and controlled creative production workflows.
Compare deployment cost, cache behaviour, model transfer and optimisation payback.
Interpret research evidence, design evaluations and assess integrated prompt programmes.

Course Content

The modern prompt engineering workbench
Establish the compact foundation needed to design advanced prompt programmes.

1.1 From a single prompt to a prompt programme
1.2 The design levers: instructions, context, compute and weights
1.3 A task contract strong enough for advanced work
1.4 Read outputs and experiments like an engineer
1.5 Studio: specify the Atlas workbench
Section 1 checkpoint: The modern prompt engineering workbench

Instruction architecture and structured control
Build instructions and output interfaces that can support complex workflows.

Advanced in-context learning and demonstration engineering
Turn examples into a deliberately designed learning signal.

Modern reasoning strategies and inference-time compute
Choose how much reasoning work to request and how to allocate it.

Inference-time search, ensembles and verification
Generate, explore and select solutions using explicit search policies.

Adaptive prompting and workflow control
Build conditional prompt programmes that change strategy as the task develops.

Automatic prompt optimisation
Move from ad hoc revisions to systematic search over prompt candidates.

DSPy, joint optimisation and GEPA
Understand prompt programming and modern feedback-driven optimisation in depth.

Agentic prompt design and tool use
Design agents that plan, use tools, inspect observations and complete bounded tasks.

Multi-agent cooperation, critique and interoperability
Coordinate specialised agents with clear communication and evidence-backed resolution.

Research synthesis and evidence-intensive prompting
Design sophisticated research and synthesis prompts over multiple sources.

Multimodal reasoning across images, audio and video
Build analysis programmes that use several media types with precise temporal and evidential control.

Creative direction and multimodal production
Direct complex creative work through reference, structure and controlled iteration.

Prompt performance, economics and cross-model transfer
Optimise deployment cost, latency and quality across different model environments.

Evaluation, robustness and the research frontier
Use rigorous experiments to advance a prompt programme and interpret new methods.

Integrated professional mastery
Compose the advanced methods into complete, purposeful AI workflows.

Add this certificate to your resume to demonstrate your skills & increase your chances of getting noticed.

Student Ratings & Reviews

No Review Yet

Prompt Engineering Mastery II: Advanced Reasoning, Agents and Prompt Optimisation

Course Prerequisite(s)

About Course

Prompt Engineering Mastery II: Advanced Reasoning, Agents and Prompt Optimisation

Go beyond a single prompt

Advanced depth, clear explanations

Build a private mastery workbench

Automatically marked assessments

A connected three-course pathway

What Will You Learn?

Course Content

The modern prompt engineering workbench Establish the compact foundation needed to design advanced prompt programmes.

1.1 From a single prompt to a prompt programme

1.2 The design levers: instructions, context, compute and weights

1.3 A task contract strong enough for advanced work

1.4 Read outputs and experiments like an engineer

1.5 Studio: specify the Atlas workbench

Section 1 checkpoint: The modern prompt engineering workbench

Instruction architecture and structured control Build instructions and output interfaces that can support complex workflows.

2.1 Layer instructions by responsibility

2.2 Design schemas and output protocols

2.3 Compose constraints and preserve invariants

2.4 Make output repair a bounded operation

2.5 Studio: build a reusable instruction stack

Section 2 checkpoint: Instruction architecture and structured control

Advanced in-context learning and demonstration engineering Turn examples into a deliberately designed learning signal.

3.1 What examples teach beyond the instruction

3.2 Select examples dynamically

3.3 Engineer contrastive and difficult examples

3.4 Order, ablate and reserve tests

3.5 Studio: construct an adaptive demonstration pack

Section 3 checkpoint: Advanced in-context learning and demonstration engineering

Modern reasoning strategies and inference-time compute Choose how much reasoning work to request and how to allocate it.

4.1 Reasoning models and the modern instruction contract

4.2 Decomposition, least-to-most and executable reasoning

4.3 Budget reasoning across easy and hard requests

4.4 Revision, critique and external feedback

4.5 Studio: allocate a reasoning budget

Section 4 checkpoint: Modern reasoning strategies and inference-time compute

Inference-time search, ensembles and verification Generate, explore and select solutions using explicit search policies.

5.1 Self-consistency and diverse candidate generation

5.2 Best-of-N and verifier-aware selection

5.3 Tree of Thoughts and bounded search

5.4 Combine partial solutions and verify the synthesis

5.5 Studio: solve the same problem three ways

Section 5 checkpoint: Inference-time search, ensembles and verification

Adaptive prompting and workflow control Build conditional prompt programmes that change strategy as the task develops.

6.1 Route by task, uncertainty and consequence

6.2 Plan, execute, inspect and replan

6.3 Build controllers with explicit states

6.4 Recursive prompting and external-input inspection

6.5 Studio: an adaptive research controller

Section 6 checkpoint: Adaptive prompting and workflow control

Automatic prompt optimisation Move from ad hoc revisions to systematic search over prompt candidates.

7.1 The optimisation loop: propose, measure, select

7.2 OPRO and score-conditioned proposal

7.3 ProTeGi and textual-gradient search

7.4 TextGrad and component-level improvement

7.5 Studio: run a prompt optimiser on paper

Section 7 checkpoint: Automatic prompt optimisation

DSPy, joint optimisation and GEPA Understand prompt programming and modern feedback-driven optimisation in depth.

8.1 Programme with signatures and modules

8.2 Bootstrap demonstrations and optimise them

8.3 MIPROv2 and SIMBA: different search mechanisms

8.4 GEPA: reflective evolution and complementary candidates

8.5 Studio: optimise a compound programme

Section 8 checkpoint: DSPy, joint optimisation and GEPA

Agentic prompt design and tool use Design agents that plan, use tools, inspect observations and complete bounded tasks.

9.1 ReAct as an executable interaction pattern

9.2 Prompt the tool interface precisely

9.3 Dependencies, parallelism and recovery

9.4 Specify initiative, permissions and completion

9.5 Studio: build an agentic research assistant

Section 9 checkpoint: Agentic prompt design and tool use

Multi-agent cooperation, critique and interoperability Coordinate specialised agents with clear communication and evidence-backed resolution.

10.1 Design a useful specialist team

10.2 Critic, verifier and judge: different jobs

10.3 Structured handoffs, MCP and A2A

10.4 Debate, diversity and triggered verification

10.5 Studio: an evidence-led specialist council

The modern prompt engineering workbench
Establish the compact foundation needed to design advanced prompt programmes.

Instruction architecture and structured control
Build instructions and output interfaces that can support complex workflows.

Advanced in-context learning and demonstration engineering
Turn examples into a deliberately designed learning signal.

Modern reasoning strategies and inference-time compute
Choose how much reasoning work to request and how to allocate it.

Inference-time search, ensembles and verification
Generate, explore and select solutions using explicit search policies.

Adaptive prompting and workflow control
Build conditional prompt programmes that change strategy as the task develops.

Automatic prompt optimisation
Move from ad hoc revisions to systematic search over prompt candidates.

DSPy, joint optimisation and GEPA
Understand prompt programming and modern feedback-driven optimisation in depth.

Agentic prompt design and tool use
Design agents that plan, use tools, inspect observations and complete bounded tasks.

Multi-agent cooperation, critique and interoperability
Coordinate specialised agents with clear communication and evidence-backed resolution.

Research synthesis and evidence-intensive prompting
Design sophisticated research and synthesis prompts over multiple sources.

Multimodal reasoning across images, audio and video
Build analysis programmes that use several media types with precise temporal and evidential control.

Creative direction and multimodal production
Direct complex creative work through reference, structure and controlled iteration.

Prompt performance, economics and cross-model transfer
Optimise deployment cost, latency and quality across different model environments.

Evaluation, robustness and the research frontier
Use rigorous experiments to advance a prompt programme and interpret new methods.

Integrated professional mastery
Compose the advanced methods into complete, purposeful AI workflows.