AI CareerPathJob board
← All projects
🔬

LLM Evaluation Platform

Advanced

2–3 weeks · 6 milestones

Build a platform for evaluating LLM outputs — RAGAS-style metrics, human evaluation UI, leaderboard, and automated regression testing.

Difficulty

Advanced

Duration

2–3 weeks

Milestones

6 steps

Interview Qs

4 questions

🛠 Tech stack

PythonRAGASFastAPINext.jsPostgreSQL or FirestoreDocker

✅ Prerequisites

  • RAG fundamentals
  • Python (advanced)
  • Full-stack basics

Step-by-step milestones

1

Define evaluation metrics

Implement faithfulness, relevance, context precision.

2

Automated eval pipeline

Run evals programmatically against a test set.

3

Human eval UI

Build side-by-side comparison and rating UI.

4

Leaderboard

Store and display model comparison results.

5

CI integration

Run evals on every PR automatically.

6

Report generation

Export evaluation reports as PDF/HTML.

Skills you'll build

LLM EvaluationRAGMetricsSystem designFrontend + Backend

📝 Interview questions you'll face

  • 1.How do you define ground truth for LLM outputs?
  • 2.What is faithfulness vs relevance in RAG eval?
  • 3.How do you prevent LLM judges from being biased?
  • 4.How would you scale evals to 1M queries/day?

🌟 Portfolio guidance

This is a standout portfolio project. Write a technical blog post, share it on LinkedIn/X, and mention RAGAS/LLM-as-judge concepts.

📚 Learn the concepts first