DATA 202 Module 11: Capstone Project

Introduction

The capstone project is the culmination of DATA 202, where you integrate everything learned throughout both courses into a substantial, original data science project. Unlike the DATA 201 project, this capstone challenges you to work with advanced data types, modern AI tools, and production-quality engineering practices.

Part 1: Capstone Expectations

What Distinguishes a Capstone

A capstone project should demonstrate:

Technical Depth: Use of advanced techniques from DATA 202

• Non-traditional data types (audio, video, documents, graphs)
• Modern ML approaches (transformers, foundation models)
• Production-quality implementation

Originality: Not a tutorial reproduction

• Novel question or application
• Creative problem framing
• Contribution beyond coursework

Completeness: Full lifecycle coverage

• Data acquisition (APIs, scraping, collection)
• Rigorous analysis and modeling
• Deployment or deployment-ready system
• Documentation and reproducibility

Reflection: Critical analysis

• What worked and what didn’t
• Limitations and future directions
• Ethical considerations

Scope Guidelines

Too Small:

• Replicating a tutorial with different data
• Simple classification on a clean dataset
• API wrapper without substantial analysis

Too Large:

• Multi-year research project
• Production system for a company
• Something requiring resources beyond your access

Just Right:

• Novel application of course techniques
• Integration of multiple data types or methods
• Working prototype demonstrating capability

Part 2: Project Categories

Category A: Novel Data Applications

Apply advanced data science to a domain or dataset not typically explored:

Examples:

• Analyze Arabic poetry with NLP techniques
• Study traffic patterns in Beirut from video feeds
• Map the network of regional art collectors
• Build a dialect classifier for Levantine Arabic

Requirements:

• Original data collection or curation
• Domain expertise or research
• Novel insights beyond obvious analysis

Category B: Multi-Modal Systems

Combine multiple data types or modalities:

Examples:

• Music video analysis (audio + video)
• Social media sentiment with image and text
• Document processing pipeline (OCR + NLP)
• Podcast transcription and topic modeling

Requirements:

• Integration of techniques from multiple modules
• Meaningful combination (not just sequential)
• Demonstration of cross-modal insights

Category C: Deployed Applications

Build a working application that serves predictions:

Examples:

• Real-time translation app for regional dialects
• Document classification service for a domain
• Recommendation system with web interface
• Monitoring dashboard with ML predictions

Requirements:

• Working deployment (local or cloud)
• API or user interface
• Documentation for use and maintenance
• Monitoring or evaluation plan

Category D: Research Replication and Extension

Replicate and extend a published paper:

Examples:

• Replicate a fairness analysis on local data
• Apply a new technique to a classic dataset
• Evaluate claims from a paper with different methods
• Combine approaches from multiple papers

Requirements:

• Complete replication (or documented failure)
• Substantive extension
• Critical analysis of original claims

Part 3: Project Phases

Phase 1: Proposal (Week 1-2)

Deliverable: 2-3 page proposal including:

• Problem statement and motivation
• Data sources and acquisition plan
• Methodology overview
• Timeline and milestones
• Potential challenges and mitigations

Review Process: Instructor feedback and approval

Phase 2: Data and Exploration (Week 3-4)

Deliverable: Progress report with:

• Acquired/collected data
• Exploratory analysis
• Refined methodology
• Any pivots from proposal

Checkpoint: Verify feasibility and scope

Phase 3: Implementation (Week 5-8)

Work: Core technical development

• Feature engineering and preprocessing
• Model development and evaluation
• Integration and deployment (if applicable)
• Iteration based on results

Mid-Point Check-in: Brief progress presentation

Phase 4: Documentation and Presentation (Week 9-10)

Deliverables:

• Final report (10-15 pages)
• Working code repository
• Presentation (15-20 minutes)
• Demo (if applicable)

Part 4: Deliverables in Detail

Final Report

Structure:

1. Abstract: One paragraph summary
2. Introduction: Problem, motivation, contributions
3. Related Work: Prior approaches and context
4. Data: Sources, collection, description, limitations
5. Methods: Approach, techniques, architecture
6. Results: Findings, evaluations, comparisons
7. Discussion: Interpretation, limitations, implications
8. Conclusion: Summary and future work
9. References: Proper citations

Quality Expectations:

• Clear, professional writing
• Appropriate visualizations
• Honest treatment of limitations
• Reproducibility information

Code Repository

Requirements:

• README with setup instructions
• requirements.txt or environment.yml
• Organized structure
• Documentation of key functions
• Example usage

Optional but Valued:

• Unit tests
• CI/CD configuration
• Docker deployment
• Interactive notebooks

Presentation

Components:

• Motivation and question
• Approach and methods
• Key results and insights
• Demo (if applicable)
• Limitations and future work
• Q&A

Delivery:

• Clear visuals (minimal text per slide)
• Narrative structure (tell a story)
• Technical depth calibrated to audience
• Time management

Part 5: Evaluation

Rubric

Excellence Markers

A-Level Work:

• Publishable or portfolio-ready
• Could be extended to real-world use
• Demonstrates mastery of course material
• Thoughtful, thorough, polished

B-Level Work:

• Solid technical execution
• Demonstrates competence
• Some limitations in depth or polish
• Room for improvement but fundamentally sound

C-Level Work:

• Meets basic requirements
• Technical issues or superficial analysis
• Incomplete components
• Limited effort beyond minimum

Part 6: Project Ideas

Data Acquisition Focus

• Build a dataset of regional news articles and analyze bias
• Create a database of restaurant reviews for Arabic cities
• Collect and process historical documents from archives

NLP and Language

• Sentiment analysis for Arabic dialects
• Arabic-English code-switching detection
• Named entity recognition for regional proper nouns
• Summarization of news in local languages

Computer Vision

• Traffic analysis from intersection cameras
• Building damage assessment from satellite imagery
• Plant disease detection for regional agriculture
• Art style classification for regional artists

Audio and Speech

• Dialect identification from speech
• Transcription improvements for accented speech
• Music genre classification for regional music
• Podcast analysis and recommendation

Networks and Graphs

• Academic collaboration networks in the region
• Trade networks between Middle Eastern countries
• Social influence mapping on regional platforms

Foundation Models

• Fine-tuning models for domain-specific tasks
• RAG system for specialized knowledge base
• Prompt engineering for specific applications
• Evaluation of model performance across languages

Deployed Systems

• Personal finance tracker with ML insights
• Study assistant with document understanding
• Local event recommendation system
• Health monitoring with sensor data

Part 7: Resources and Support

Office Hours

• Weekly scheduled hours for project consultation
• Additional appointments by request
• Peer feedback sessions

Computing Resources

• Cloud credits available for deployment
• GPU access for model training
• Data storage allocation

Ethical Review

• Projects involving human subjects need ethics consideration
• Discuss sensitive data with instructor
• Privacy and consent requirements

Final Thoughts

The capstone is your opportunity to synthesize everything you’ve learned and create something meaningful. The best projects come from genuine curiosity—questions you actually want to answer, problems you want to solve, tools you want to build.

Start early. Iterate often. Ask for help when stuck. And remember: the goal is learning, not perfection. A project that encountered challenges and documented them honestly is more valuable than one that hid difficulties behind polished output.

Welcome to the final challenge of DATA 202. Make it count.

Module 11 structures the capstone project—the culminating experience of DATA 202 where students integrate advanced techniques into substantial, original work. From proposal to presentation, the capstone demonstrates mastery of modern data science.