Module 18 of 18
Advanced Capstone: Production Prompt System
Design, build, test, and document a production-grade prompt system integrating techniques from the entire 18-module curriculum
Project Options
The advanced capstone is your opportunity to demonstrate mastery of the entire 18-module curriculum by designing, building, testing, and documenting a production-grade prompt system. Choose a project that aligns with your interests and career goals while challenging you to integrate techniques from across the course.
Definition
Advanced capstone
Each project option below is designed to require integration of multiple course modules. Choose the one that best aligns with your domain expertise and career interests.
AI Content Pipeline
Build an end-to-end content production system that researches topics, generates drafts, edits for quality, optimizes for SEO, and adapts content across formats (blog, social, email). The pipeline should handle the full content lifecycle from ideation to publication-ready output. Key modules: Chaining, RAG, Testing, AI Products.
Intelligent Document Processing
Create a system that ingests documents (contracts, reports, research papers), extracts structured information, classifies content, identifies anomalies, and generates actionable summaries. Handle multiple document types and formats. Key modules: RAG, Advanced Techniques, Evaluation, Governance.
Multi-Agent Customer Service
Design a customer service platform where specialized AI agents handle different request types (billing, technical, sales, escalation). Agents collaborate, hand off conversations, and maintain context across interactions. Key modules: Role Systems, Chaining, PromptOps, Testing.
Domain Research Assistant
Build a research assistant for a specific domain (medical, legal, financial, academic) that can search knowledge bases, synthesize findings, generate reports, and answer follow-up questions with source citations. Key modules: RAG, Domain Playbooks, Multimodal, Security.
AI Code Review System
Create a code review system that analyzes pull requests for bugs, security vulnerabilities, style violations, and architectural concerns. Generate actionable feedback with specific suggestions and code examples. Key modules: Model-Specific, Testing, Chaining, AI Products.
Custom Project
Propose your own project that meets the required integration criteria. Custom projects must demonstrate equivalent complexity and module integration as the predefined options. Submit a proposal for approval before starting. Must integrate 5+ modules.
Required Integrations
Regardless of which project you choose, your capstone must demonstrate proficiency in techniques from across the curriculum. The following integration requirements ensure your project exercises the full breadth of prompt engineering skills covered in this course.
Definition
Required integrations
Deliverables & Rubric
Your capstone will be evaluated across five dimensions that reflect the complete prompt engineering skill set. The rubric is designed to reward both technical depth and practical utility.
Definition
Evaluation formula
Technical Sophistication (30%)
Quality and complexity of the prompt architecture. Evaluated on prompt design, technique integration, model selection, and system design decisions. Excellent (27-30): novel techniques, optimal architecture. Good (21-26): solid design, effective technique use.
Practical Utility (25%)
Real-world usefulness of the system. Evaluated on problem significance, user value, deployment feasibility, and measurable impact. Excellent (23-25): clear value, deployment-ready. Good (18-22): useful system, some gaps.
Testing & Evaluation (20%)
Rigor of testing and evaluation approach. Evaluated on test coverage, metric selection, debugging documentation, and continuous evaluation plan. Excellent (18-20): comprehensive suite, automated metrics.
Documentation (15%)
Quality and completeness of project documentation. Evaluated on architecture docs, decision logs, setup guides, and governance documentation. Excellent (14-15): complete, clear, production-quality.
Presentation (10%)
Ability to present and defend the project. Evaluated on clarity of presentation, demo quality, handling of questions, and articulation of trade-offs and lessons learned.
Project Milestones
The capstone project follows a 4-week structure designed to ensure steady progress and provide checkpoints for feedback. Each week has specific deliverables that build toward the final submission.
- 01
Week 1: Design & Proposal
Select your project, write the proposal, and design the architecture. Deliverables include the project proposal document, architecture diagram, and initial prompt drafts for the core pipeline. Submit the proposal for peer feedback. Checkpoint: proposal reviewed, architecture approved, development environment ready. - 02
Week 2: Core Build
Build the core prompt system with all 5+ interconnected prompts. Implement the primary workflow from input to output. Begin building the test suite with at least 10 test cases. Focus on getting the happy path working end-to-end before handling edge cases. Checkpoint: core pipeline functional, 10+ test cases passing, demo-ready happy path. - 03
Week 3: Harden & Test
Expand the test suite to 20+ cases including edge cases and adversarial tests. Add error handling, fallback prompts, and security measures. Implement evaluation metrics and begin optimization. Document governance considerations and cost analysis. Checkpoint: 20+ test cases, error handling complete, evaluation metrics running, documentation in progress. - 04
Week 4: Document & Present
Complete all documentation including architecture docs, decision logs, governance documentation, and the project README. Prepare the presentation and demo. Conduct peer reviews of other projects. Polish and submit the final deliverable package. Checkpoint: all documentation complete, presentation rehearsed, peer reviews submitted, final package delivered.
The capstone is not just about building something that works. It is about demonstrating the full spectrum of prompt engineering skills: design, implementation, testing, governance, documentation, and communication. Approach it as if you were delivering a real project to a real stakeholder.
Activities
Activity 1
60-90 min · Advanced
Project Proposal
Write a project proposal including problem statement, approach, architecture diagram, and success criteria.
Activity 2
3-4 hours · Advanced
Core Prompt System Build
Build the core prompt system with at least 5 interconnected prompts. Test with 20+ cases.
Activity 3
45-60 min · Advanced
Peer Review
Conduct a peer review of another learner's project using the evaluation rubric. Provide actionable feedback.
Real-World Applications
Content Pipelines
End-to-end content production systems that automate research, drafting, editing, and distribution across multiple channels and formats.
Document Processing
Intelligent systems that extract, classify, and summarize information from unstructured documents at scale for legal, financial, and healthcare applications.
Multi-Agent Services
Customer service platforms with specialized AI agents that collaborate to handle complex multi-turn conversations with appropriate escalation paths.
Research Assistants
Domain-specific research tools that search, synthesize, and present findings with source citations and confidence indicators.
Code Review Systems
Automated code analysis tools that identify bugs, vulnerabilities, and architectural concerns while generating actionable feedback for developers.
Portfolio Showcase
The capstone project itself serves as a powerful portfolio piece demonstrating comprehensive prompt engineering competency to potential employers or clients.