Skip to content

DEMO_DAY.md

Latest commit

 

History

History
68 lines (46 loc) · 3.31 KB

DEMO_DAY.md

File metadata and controls

68 lines (46 loc) · 3.31 KB
Raw

Demo Day — XJTLU ENT 208

Context

Course: ENT 208 — Entrepreneurship and Innovation
Institution: Xi'an Jiaotong-Liverpool University (XJTLU)
Session: 2024–2025, Session 2, Group 20
Project: Twelfth Night (developed as Luckie-Bot)

The Challenge

ENT 208 is a project-based entrepreneurship course requiring student teams to:

  1. Identify a product opportunity
  2. Design and build a functional MVP
  3. Develop a credible business model
  4. Present a live demonstration to a panel of faculty and industry judges

The course emphasises real implementation over theoretical business plans. Teams must ship working products.

Our Approach

We chose to build an embodied AI companion — a product that combines generative AI, computer vision, and IoT hardware into a single coherent experience. This was deliberately ambitious for a course project:

  • Three independent technology domains (LLM, CV, embedded systems) had to work together in real time
  • Live demonstration risk was high — any single layer failure would break the entire experience
  • Hardware dependency meant we couldn't just show slides; the device had to work on stage

Our strategy was to build a robust, layered architecture where each component could degrade gracefully. The web app works standalone. The hardware enhances but is not required. The AI response has a fallback path.

The Demonstration

The live demo flow:

  1. Category Selection — Judge selects a life category on the hardware device or web interface
  2. Gesture Interaction — Cards are drawn through hand gestures tracked by MediaPipe
  3. AI Generation — DeepSeek-V3 generates a personalised response during a breathing animation
  4. Hardware Feedback — The M5StickC Plus device synchronises LED lighting with the experience
  5. Result Reveal — The AI-generated reading is displayed with visual effects
  6. Business Presentation — Product vision, market positioning, revenue model

The entire demo ran without technical failure. Real-time AI + gesture + hardware sync worked on stage.

Result

🥇 First Place — ENT 208 Demo Day

The project was evaluated across five dimensions:

Dimension Assessment
Innovation Novel intersection of AI, IoT, and wellness
Technical Depth Production-quality three-layer architecture
User Experience Polished multi-sensory interaction flow
Business Viability Clear monetisation path with defensible moat
Presentation Confident live demonstration under pressure

Key Takeaways

  1. Build for reliability, not just novelty. The live demo succeeded because we designed for failure modes at every layer.
  2. AI products need interaction design. The LLM is only one component — how users access and experience it matters equally.
  3. Hardware-software integration is a genuine differentiator. The physical companion device created an impression that a pure software demo could not.
  4. Storytelling matters in technical work. The product narrative made the technology accessible and memorable to non-technical judges.
  5. Entrepreneurship is about shipping. A working MVP with flaws beats a perfect slide deck every time.

Twelfth Night was originally developed as "Luckie-Bot" for XJTLU ENT 208 Demo Day, where it won First Place.