Overview

By the end of the course, you must develop a prototype of an AI application that addresses an important business problem. The goal for this assignment is to integrate across the various learning objectives of the class, challenging you to think holistically about how AI can solve key business problems. A few remarks about the design of the assignment:

• The assignment is intentionally open ended, so that you can choose a problem that is fun for you to work on. Choose a problem that you think is important or valuable. Spend time up front carefully considering your options!
• The project is intended to be done over a period of time, so that you have time to explore the software libraries introduced in class. Please don’t leave this to the last minute, or you will miss out an opportunity to learn these powerful technologies.
• My intention is for you to develop your project using generative AI technologies, as that is what we covered in the class. However, if you have prior experience with predictive AI technology and feel passionate about a project based on that approach, you are welcome to pursue that approach as well. Please be clear about your intention to do this in your initial proposal.
• Final project work can be done in teams of up to three; grading standards will be commensurate to the number of teammates (a higher standard for more people).
  • If you can form a larger team, I encourage you to do so. You will have a more enjoyable project experience and implement a more complete prototype.
• You are responsible for sourcing your own dataset.
  • You may use public available data from Kaggle or other common data science sources.
  • You may collect the data yourself
  • You may use data that you have access to through other activities, assuming that you have permission to do so.
• The use of generative AI is highly encouraged in the process of developing these prototypes. [See below for the course’s AI policy.] We will only have time to engage libraries like OpenAI LangChain at a superficial level in class, but you will likely need them to implement interesting solutions. This project is your chance to learn how to work with these technologies at a deeper level.

Sample Projects

[just to give you a sense of what is possible — not intended to limit you]

• Generative AI
  • Customer Support Chatbot: Build an AI-driven chatbot for automating customer support using a pre-trained language model. Show how it reduces operational costs by handling common inquiries. (Service Requests)
  • Automated Resume Screening Tool: Create a resume screening tool using natural language processing (NLP) techniques. Showcase its ability to speed up the hiring process and improve candidate selection.
  • AI-Based Recommendation System: Build a recommendation system for an e-commerce site, demonstrating how it enhances user experience and increases sales by suggesting relevant products. (Product Reviews)
  • Data Analysis Assistant: Integrate custom information into a chatbot that can facilitate business analysis.
• Prediction-based AI
  • Demand Forecasting System: Develop a system that uses historical sales data to predict future demand, demonstrating how it can reduce inventory costs and optimize supply chain management.
  • Fraud Detection in Transactions: Use AI to detect fraudulent transactions in financial data, explaining how it improves security and reduces financial losses for a business.
  • Predictive Maintenance for Machinery: Implement a predictive maintenance system using sensor data to predict equipment failure. Demonstrate how it can save costs by reducing downtime and repair expenses.

Grading

Final projects will be graded based on five components:

• Project Proposal / Description [5]. Introduce your project to a general audience. Describe what it does, how it does it, and why it’s valuable. What features do you hope your project to illustrate? First draft of proposal due February 18, 2025, and feedback will be provided by the following week.

  • For full credit, your proposal must:
    • Clearly explains the project, its purpose, and its value in a concise and compelling way.
    • Be well-defined, achievable, and appropriately ambitious given the time constraints of the class.
  • In certain situations, the instructor may ask for more detail or ask for a more ambitious implementation.

• Business Problem Framing (750 words max) [10]. What business problem are you solving? How does AI fit into your solution? How will your solution impact the underlying fundamentals of the business? First draft of problem framing due March 4, 2025, and feedback will be provided.

  • For full credit, your problem framing must:
    • Clearly defines a relevant business problem with strong supporting context.
    • Demonstrates a clear and logical application of AI in solving the problem.
    • Provides strong analysis of the business impact, supported by examples or data.

• Organizational Implementation Plan (750 words max) [10]. How will you evaluate your solution? What ethical, privacy, or bias-related issues could emerge? What additional work would be needed to implement a solution like this in practice in an organization? First draft of problem framing due March 25, 2025, and feedback will be provided.

  • For full credit, your implementation plan must
    • Clearly defines how success will be measured with strong, relevant metrics.
    • Thoroughly addresses bias, privacy, and ethical risks with mitigation strategies.
    • Discuss real-world implementation challenges and solutions, connecting them to themes and examples covered in the class.

• Working Technical Prototype [10]. You must implement a prototype of your proposed solution.

  • You may use frameworks such as Gradio, Hugging Face Spaces, or any platform where you can showcase the application in action.

  • For full credit, your prototype must

    • be fully functional. You must submit one of the following:
      • A colab notebook that runs end-to-end
      • Your code + a working Gradio example accessible via link (if you wish to use proprietary resources)
      • Your code + clear instructions for how to install and run your code. Your code must be usable on a standard computing setup (no GPUs assumed).
    • demonstrating mastery of the core AI features covered in the class.
    • provide evidence independent exploration of the openai, langchain, langgraph, gradio, or related machine-learning / AI libraries.
      • implement at least $N$ non-trivial feature not covered in class materials using these libraries, where $N$ is the number of project participants.

    Alternatively, you are welcome to develop a ML model and build it into some user-facing product using alternative machine learning libraries. (Please consult the instructor before doing this, given that it will involve technical approaches not covered in this class).

• Presentation [5]. You will create an engaging video presentation demoing your project (likely around 2-5 minutes). The top projects may have an opportunity to present their project live for the class with Q&A for extra credit opportunities.

  • For full credit, your presentation must be:
    • Well-structured and clear.
    • Engaging delivery with strong visuals and motivation.
    • Clearly demonstrates how the prototype works with clear and concise explanations.

Final Project Submission Instructions