Students can choose one of the following two tracks:

1. Application Project (Build)

• Create a system using LLMs, diffusion models, or agents.
• Examples:
  • A text-to-image web app (Stable Diffusion + Streamlit)
  • A chatbot that retrieves and summarizes scientific articles (RAG + LLM)
  • A multi-agent simulation (LangGraph / CAMEL)
  • A music generator using transformer models

2. Analysis Project (Investigate)

• Study a GenAI model’s behavior or risks.
• Examples:
  • Measure hallucination rates in GPT-3 across domains
  • Quantify bias in Stable Diffusion prompts
  • Explore privacy leakage (membership inference on GPT-2)
  • Robustness analysis: how easily can an LLM be jailbroken?

Deliverables

• A 4-6 page technical report, including introduction, methods, results, and discussions.
• The complete code package.
• Details