Software Developer Intern

Create an AI Judge

1. Project Overview
a. Side A – Lawyer enters documents in the form of text, PDF, Word, and other case-related information to the AI.
b. Side B – Does the same.
c. Outcome – Our AI, trained on judgments from a particular country (with plans to handle international cases as well), gives its verdict. This functions like a mock trial outcome.
d. Next – After the initial decision, the sides can argue the case and suggest alternative approaches (similar to asking Claude, “Have you thought about XYZ way of doing this? That would be better,” and the AI responds, “Yes, that is a better way. Let me reconsider.”). The AI will then reevaluate the pros and cons.
e. Interface – After argument and cross-questioning, there will be a clear distinction on the screen: a central judge icon and left/right chat boxes for both sides.

2. Constraints
a. A limited number of follow-up arguments after the AI decides (maximum of 5).

3. Originality
a. What additional features can be added to improve this AI Judge?
b. What innovative approaches can you think of to productize this system?
 

Project Notes

• Use any LLM API: ChatGPT, Claude, Gemini, etc.
• Use any programming language: Python, JavaScript, TypeScript, Go (avoid Java and PHP for backend).
• This internship is not just about writing code—it’s about how you think, craft, and bring your ideas to life through software.
• Use any coding assistant you prefer: Claude Code, Cursor, Codex, etc.
• You define the project architecture: whether this is frontend-only or requires a backend is up to you.

Evaluation Criteria for Interview

• UI/UX: How the interface looks and how easily (yet meaningfully) the user can use the system.
• Product Strategy: The rationale behind every element—why a button exists, why an extra field exists, etc.
• Code Quality: Modularity, structure, and function choices.
• Scalability: How the system can handle thousands of users.
• Caching Strategy: How data and computations will be cached efficiently.
• Deployment: Tools and strategies for production deployment (e.g., Kubernetes, Ansible, telemetry, secrets manager, AWS, Azure, GCP).
• Impact: How this system can benefit the judicial system in India or other countries.

How did you approach designing the AI Judge project? How did you choose the tech stack, and why?

Why did you choose MongoDB over MySQL or PostgreSQL, given that this project involves relationships between two parties and previous judgments?

You avoided using the useEffect hook and used TanStack instead. Can you explain the benefits it offers compared to making API calls within useEffect?

How did you implement the AI Judge functionality? Did you use a trained model, and how did you decide which country’s laws to base the judgments on?

How can you hide hydration warnings and errors in Next.js?

What is the purpose of a sitemap and RSS feeds in SEO?

Quickly create an Express server, connect it to any LLM API model, and create a simple chat endpoint.