AI/Machine Learning Intern Challenge: Simple Content-Based Recommendation

Deadline: Sunday, Feb 23th 11:59 pm PST

---

Overview

Build a content-based recommendation system that, given a short text description of a user’s preferences, suggests similar items (e.g., movies) from a small dataset. This challenge should take about 3 hours, so keep your solution simple yet functional.

Example Use Case

• The user inputs:
  "I love thrilling action movies set in space, with a comedic twist."
• Your system processes this description (query) and compares it to a dataset of items (e.g., movies with their plot summaries or keywords).
• You then return the top 3–5 “closest” matches to the user.

---

Requirements

1. Dataset

   • Use a small public dataset of items (e.g., a list of movies with plot summaries, or other textual descriptions).
   • Make sure the dataset is easy to handle (maybe 100–500 rows) so the solution remains quick to implement and run.
   • Include the dataset in your forked repository or provide instructions/link on how to download it.

2. Approach

   • Content-Based: At a minimum, use text similarity to recommend items.
     • For instance, you can transform both the user’s text input and each item’s description into TF-IDF vectors and compute cosine similarity.
   • Return the top N similar items (e.g., top 5).

3. Code Organization

   • You may use a Jupyter Notebook or Python scripts.
   • Keep it readable and modular (e.g., one section for loading data, one for building vectors, one for computing similarity, etc.).
   • Briefly comment or docstring your key functions/sections.

4. Output

   • When given an input description (e.g., "I like action movies set in space"), your system should print or return a list of recommended items (e.g., 3–5 titles).
   • Include the similarity score or rank if you’d like.

5. Summary & Instructions

   • A short README.md that includes:
     • Dataset: Where it’s from, any steps to load it.
     • Setup: Python version, virtual environment instructions, and how to install dependencies (pip install -r requirements.txt).
     • Running: How to run your code (e.g., python recommend.py "Some user description" or open your notebook in Jupyter).
     • Results: A brief example of your system’s output for a sample query.

---

Deliverables

1. Fork the Public Repository

   • Fork this repo into your own GitHub account.

2. Implement Your Solution

   • Load and preprocess your dataset (e.g., read CSV, handle text columns).
   • Convert text data to vectors (e.g., TF-IDF).
   • Implement a function to compute similarity between the user’s query and each item’s description.
   • Return the top matches.
   • Salary expectation per month (Mandatory)

3. Short Video Demo

   • In a .md file (e.g., demo.md) within your fork, paste a link to a brief screen recording (video link).
   • Demonstrate:
     • How you run the recommendation code.
     • A sample query and the results.

4. Deadline

   • Submit your fork by Sunday, Feb 23th 11:59 pm PST.

Note: This should be doable within ~3 hours. Keep it straightforward—you do not need advanced neural networks or complex pipelines. A simple TF-IDF + cosine similarity approach is sufficient.

---

Evaluation Criteria

1. Functionality

   • Does your code run without errors?
   • When given an input query, does it successfully output relevant items?

2. Code Quality

   • Clear, commented code (where it counts).
   • Logical steps (load data → transform → recommend).

3. Clarity

   • Is your README.md straightforward about setup, how to run, and what to expect?

4. ML/Recommendation Understanding

   • Basic implementation of a content-based recommendation approach (vectorization, similarity measure).

We look forward to seeing your solution! Good luck!