Final Project: Proposal

Due: Tuesday, March 24 @ 11:59 PM ET (submit on Gradescope)

Overview

For the final project, you will work in groups of 2 to design and build an LLM agent that solves a specific, well-scoped problem. Your proposal should describe the problem you want to solve and how your agent will be structured at a high level.

Submit a 2-page PDF proposal document that clearly explains the agent you plan to build.

What Your Proposal Should Cover

Your 2-page proposal document should include the following:

1. Problem, motivation, and scope

Explain the problem your agent will solve and why it matters. State the intended user or user group. Define exactly what the agent will do. Your project should be narrow enough to complete in one month.

2. High-level agent design

Describe, at a high level, how your agent will work. You do not need full implementation details yet, but you should identify the major components your final system is expected to have, including:

  • Data collection and parsing
  • A memory store + retrieval mechanism
  • At least 3 tools the agent can call
  • Decision logic or system flow
  • Guardrails or safety considerations

3. Implementation and evaluation plan

Briefly explain how you will tell whether the project succeeds. What kinds of tasks, examples, or failure cases will you test? What challenges do you anticipate running into and how will you address them?

Examples of Well-Scoped Projects

  • A travel planning agent that can help users plan their trips to a specific destination for a specific date range, budget, and preferences. The agent can be integrated with various external tools to handle user requests and book flights, hotels, and activities in the destination.
  • An apartment-shortlisting agent for someone looking to rent an apartment in a specific city. The agent can dynamically update available apartments based on user preferences and constraints, like safety or commute time, and contact landlords or property managers to schedule viewings.
  • An environmental analysis agent that can help business owners analyze the environmental impact of their company. The agent can review documents and analyze production pipelines to measure carbon emissions, electricity, or water usage, and offer suggestions for reducing environmental impact without profit loss.