- Dates: March 9 - March 20, 2026 (4 sessions)
- Format: Group presentations, 20 minutes per group
- Topics: 4, with increasing difficulty
| Session | Date | Topic | Difficulty |
|---|---|---|---|
| #1 | 09/03 | Startup Toolbox | Beginner |
| #2 | 13/03 | Claude Code Deep Dive | Intermediate |
| #3 | 19/03 | Attention Is All You Need | Advanced |
| #4 | 20/03 | OpenClaw & Agent Ecosystem | Expert |
Later topics are more difficult, but come with more preparation time.
The following are reference directions, not a checklist to cover exhaustively. Each group should choose their focus based on their own understanding — depth matters more than breadth.
Preparing for the Demo Day NLP project, sharing practical tools and experience:
- Project management (e.g., Linear, GitHub Projects)
- Getting free LLM API tokens
- Vibe Coding tools
- Frontend deployment (e.g., Vercel)
- Backend hosting
Mastering Claude Code's advanced features and workflows:
- MCP (Model Context Protocol)
- Creating and using Skills
- Prompt engineering & CLAUDE.md
- Tool Use techniques
- Terminal environments (e.g., Conductor, tmux)
Understanding the core of the Transformer architecture:
- Self-Attention mechanism (Query, Key, Value)
- Multi-Head Attention
- Positional Encoding
- Encoder-Decoder architecture
- From Transformer to BERT, GPT
Understanding AI Agent design and development trends:
- OpenClaw configuration and usage
- OpenClaw's design philosophy and architecture
- Platform ecosystem (e.g., Moltbook)
- Agent-to-Agent (A2A) protocols and trends
This is not a traditional academic presentation — it is a knowledge sharing and learning reflection exercise.
| Dimension | Description | Weight |
|---|---|---|
| Knowledge Transfer | Did the audience actually learn something? | 40% |
| Learning Process & Methods | Did the group show how they learned the topic? (AI tools used, peer discussions, challenges encountered) | 40% |
| Presentation Quality | Clear delivery, good structure, time kept within 20 minutes | 20% |