Presentation

Due: Week of April 20, 2026 (schedule TBA) Points: 150 Format: 10-minute presentation + 5-minute Q&A

Overview

The final presentation showcases your project to the class. You will present your research question, demonstrate your system, and discuss your findings. This is an opportunity to communicate your work to a broader audience and practice presenting technical research.

Deliverables

Live Presentation - 10-minute in-class presentation
Demo - Working demonstration of your system
Slides - Presentation slides (PDF in repository)
Video (optional) - Pre-recorded backup in case of technical issues

File Structure

cis6930sp26-project/
├── presentation/
│   ├── slides.pdf
│   └── demo_notes.md (optional)
└── ...

Presentation Structure

Recommended Outline (10 minutes)

Section	Time	Content
Introduction	1-2 min	Problem, motivation, research question
Approach	2-3 min	System architecture and key design decisions
Demo	2-3 min	Live demonstration of working system
Results	2-3 min	Key findings with evidence
Conclusion	1 min	Summary and takeaways

What to Cover

Introduction (1-2 minutes)

Hook: Why should the audience care?
Problem statement: What challenge are you addressing?
Research question: What specifically did you investigate?

Approach (2-3 minutes)

Architecture overview: High-level system diagram
Key components: MCP servers, LLM orchestration
Design decisions: Why these choices?

Demo (2-3 minutes)

Show the system working
Walk through a typical use case
Highlight LLM interactions

Results (2-3 minutes)

Main findings with specific numbers
Comparison to baselines
What surprised you?

Conclusion (1 minute)

Summary of contributions
Limitations and future work
One takeaway for the audience

Rubric

Criterion	Weight	Points	Description
Content	30%	45	Does the presentation cover essential elements?
Clarity	25%	37.5	Is the presentation easy to follow?
Demo	20%	30	Is the demonstration effective?
Delivery	15%	22.5	Is the presentation professionally delivered?
Q&A	10%	15	Are questions handled well?
Total	100%	150

Scoring Scale

Score	Meaning
5	Excellent - Conference-quality presentation
4	Good - Professional and engaging
3	Satisfactory - Gets the message across
2	Below Average - Hard to follow or missing elements
1	Poor - Does not meet basic requirements

Detailed Criteria

Content (30%)

Score	Description
5	Comprehensive coverage; clear problem, approach, and results; insights shared
4	Good coverage of all key elements
3	Covers basics but missing depth in some areas
2	Incomplete coverage; key elements missing
1	Does not adequately present the work

Required Elements:

Problem statement and motivation
Approach/methodology
Working demonstration
Key results and findings
Conclusions

Clarity (25%)

Score	Description
5	Crystal clear; excellent logical flow; complex ideas explained well
4	Clear and well-organized
3	Understandable but some confusing parts
2	Difficult to follow; organizational issues
1	Incomprehensible

Guiding Questions:

Can someone unfamiliar with the project understand it?
Is the presentation logically organized?
Are technical concepts explained at an appropriate level?

Demo (20%)

Score	Description
5	Compelling demo; shows system working; highlights key capabilities
4	Good demo that supports the presentation
3	Demo present but could be more effective
2	Demo has issues or doesn’t illustrate key points
1	No demo or demo fails completely

Demo Requirements:

Show the system running
Demonstrate MCP server interactions
Show LLM orchestration in action
Highlight interesting behaviors

Delivery (15%)

Score	Description
5	Confident, engaging, professional
4	Good delivery with clear communication
3	Adequate delivery; some nervousness or pacing issues
2	Difficult to hear or understand; poor pacing
1	Unprepared or unprofessional

Guiding Questions:

Is the speaker audible and clear?
Is the pace appropriate?
Does the speaker engage the audience?
Is the speaker prepared?

Q&A (10%)

Score	Description
5	Handles all questions confidently; provides insightful answers
4	Good responses to questions
3	Answers most questions adequately
2	Struggles with questions; incomplete answers
1	Cannot answer basic questions

Slide Examples

Title Slide

┌────────────────────────────────────────────────────┐
│                                                    │
│    TransitLLM: LLM-Orchestrated Smart City         │
│    Data Integration                                │
│                                                    │
│    Your Name                                       │
│    CIS 6930 - Spring 2026                          │
│                                                    │
└────────────────────────────────────────────────────┘

Problem Slide

┌────────────────────────────────────────────────────┐
│  The Problem                                       │
│                                                    │
│  Smart city data is fragmented across portals      │
│                                                    │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐            │
│  │ Transit │  │Utilities│  │   311   │            │
│  │  API    │  │   API   │  │   API   │            │
│  └────┬────┘  └────┬────┘  └────┬────┘            │
│       │            │            │                  │
│       ▼            ▼            ▼                  │
│  Different schemas, formats, quality levels        │
│                                                    │
│  → Manual ETL is time-consuming and error-prone   │
│                                                    │
└────────────────────────────────────────────────────┘

Architecture Slide

┌────────────────────────────────────────────────────┐
│  System Architecture                               │
│                                                    │
│  ┌──────────────────────────────────────────┐     │
│  │           LLM Orchestrator               │     │
│  │  (planning, schema mapping, validation)  │     │
│  └────────────────┬─────────────────────────┘     │
│                   │                                │
│      ┌────────────┼────────────┐                  │
│      ▼            ▼            ▼                  │
│  ┌───────┐   ┌───────┐   ┌───────┐               │
│  │Transit│   │Utility│   │  311  │               │
│  │ MCP   │   │ MCP   │   │ MCP   │               │
│  └───────┘   └───────┘   └───────┘               │
│                                                    │
└────────────────────────────────────────────────────┘

Results Slide

┌────────────────────────────────────────────────────┐
│  Key Results                                       │
│                                                    │
│  Schema Mapping Accuracy                           │
│  ┌────────────────────────────────────────┐       │
│  │ Baseline (manual)  ████████████████ 96% │       │
│  │ TransitLLM        ███████████████ 94%  │       │
│  └────────────────────────────────────────┘       │
│                                                    │
│  Development Time (hours)                          │
│  ┌────────────────────────────────────────┐       │
│  │ Baseline (manual)  ████████████████ 16h │       │
│  │ TransitLLM        ██████ 6h             │       │
│  └────────────────────────────────────────┘       │
│                                                    │
│  → 60% reduction in development time              │
│  → Comparable accuracy                            │
│                                                    │
└────────────────────────────────────────────────────┘

Demo Tips

Before the Presentation

Test everything - Run through the demo multiple times
Prepare fallbacks - Have screenshots or video backup
Clear your screen - Hide notifications, bookmarks, personal info
Check network - Ensure APIs and LLMs are accessible
Pre-load data - Avoid long waits during the demo

During the Demo

Narrate your actions - Explain what you’re doing
Highlight key moments - Point out LLM decisions
Keep it focused - Show the most interesting parts
Handle errors gracefully - Have a plan if something fails
Stay calm - Small issues happen; move forward

Demo Script Example

“Let me show you the system in action. I’m going to query the transit MCP server for today’s bus routes.”

Runs command

“You can see the LLM is now orchestrating the extraction. It’s detected that the transit API uses a different date format than our target schema.”

Points to output

“Here the LLM has automatically applied the correct transformation and validated the data quality.”

Q&A Preparation

Anticipate Questions

Prepare answers for:

Why did you choose this dataset/approach?
How does this compare to [alternative approach]?
What happens when [edge case]?
What would you do differently?
How could this be extended?

Handling Difficult Questions

Take a moment - It’s okay to pause and think
Clarify if needed - “Do you mean…?”
Be honest - “I didn’t test that, but I would expect…”
Stay focused - Don’t go on tangents
Offer to follow up - “I can look into that and get back to you”

Common Mistakes to Avoid

Too much text on slides - Use bullet points, not paragraphs
Reading from slides - Slides support your talk, they don’t replace it
Skipping the motivation - Explain why anyone should care
Demo failures without backup - Always have screenshots/video
Running over time - Practice timing and respect the limit
Ignoring the audience - Make eye contact, don’t stare at screen
Rushing through results - This is the payoff; give it time

Submission Checklist

Slides are complete and polished
Demo is tested and working
Backup demo (screenshots/video) is prepared
Presentation is timed (10 minutes)
Q&A answers are prepared
Slides PDF is in presentation/ directory
Presentation date is confirmed

Resources

Project Overview - Full project description
Final Paper - Previous milestone
Video Rubric - For video presentations

back