Scribd
Upload
4 views2 pages

Intern Assignment

The assignment involves creating a 1-Dimension Adaptive Testing Prototype to enhance student assessments through AI. It requires designing a MongoDB database with GRE-style questions, implementing an adaptive algorithm in Python, and optionally integrating an LLM for personalized learning insights. Deliverables include source code, a README with project instructions, and API documentation, with evaluation based on system design, algorithm logic, AI proficiency, and code quality.

Uploaded by

Prashant Singh
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views2 pages

Intern Assignment

The assignment involves creating a 1-Dimension Adaptive Testing Prototype to enhance student assessments through AI. It requires designing a MongoDB database with GRE-style questions, implementing an adaptive algorithm in Python, and optionally integrating an LLM for personalized learning insights. Deliverables include source code, a README with project instructions, and API documentation, with evaluation based on system design, algorithm logic, AI proficiency, and code quality.

Uploaded by

Prashant Singh
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Intern Assignment: AI-Driven Adaptive Diagnostic Engine

This is a professional-grade assignment description designed to filter for architectural


thinking and AI-augmented productivity.

Overview
At our core, we aim to revolutionize how students learn by making assessments smarter.
Your task is to build a 1-Dimension Adaptive Testing Prototype. This system should
determine a student's proficiency level by dynamically selecting questions based on their
previous answers.

Tools Allowed
We encourage the use of [Link], ChatGPT, Claude, or GitHub Copilot. We aren't testing
your ability to memorize syntax; we are testing your ability to build robust, scalable AI
systems.

Tech Stack Requirements:


* Database: MongoDB (Atlas or Local)

* Language: Python (FastAPI or Flask preferred for the backend)

* AI Integration: OpenAI API or Anthropic API

* Architecture: Clean, modular code with clear separation of concerns.

The Task:
Phase 1: Data Modeling (MongoDB)
Design and seed a MongoDB collection of at least 20 GRE-style questions.

* Each question must have a `difficulty` score (0.1 to 1.0).

* Include metadata: `topic` (e.g., Algebra, Vocabulary), `tags`, and `correct_answer`.

* Create a `UserSession` collection to track a student's progress and current estimated


ability.

Phase 2: The Adaptive Algorithm


Implement a 1D Adaptive Logic in Python.

1. Starting Point: The student begins at a baseline difficulty of 0.5.


2. The Loop: If the student gets a question correct, the next question should be harder. If
incorrect, the next question should be easier.

3. Refinement: Use IRT (Item Response Theory) approach to update the user’s "Ability
Score" after every response.

Phase 3: AI Insights (The bonus)


Integrate an LLM to provide a Personalized Learning Plan.

* Once the test ends (e.g., after 10 questions), send the student's performance data (topics
missed, difficulty level reached) to an LLM.

* Generate a 3-step study plan tailored to their specific weaknesses.

Deliverables:
1. Source Code: A GitHub repository link.

2. [Link]:

* Instructions on how to run the project.

* A brief explanation of your adaptive algorithm logic.

* The "AI Log": A short section describing how you used AI tools (Cursor/ChatGPT) to speed
up your development and any challenges the AI couldn't solve.

3. API Documentation: A simple list of endpoints (e.g., `GET /next-question`, `POST


/submit-answer`).

Evaluation Criteria
Criteria & What we are looking for:

1. System Design - How you structured the MongoDB schema for performance.

2. Algorithmic Logic - Is your adaptive logic mathematically sound or just a random jump?

3. AI Proficiency - Efficient use of LLM prompts for generating the study plan.

4. Code Hygiene - Proper error handling, environment variable usage, and typing.

> We value quality over quantity. A perfectly functioning 5-question demo with a solid
algorithm and clean code is better than a broken 50-question system.

You might also like