Interview Twin: Dveloping a Production-Ready LLM Application

Architecture Diagram

(to be added as the project evolves)

Why

1. Personal helper for interview preparation
2. To demonstrate the process of developing a full end-to-end product
3. Networking, self-advertising, and learning in public

What

System scope:

• The goal is to Production-ready and deployable on the cloud — not just a demo.

References

• LLM Engineer’s Handbook
• LLM Zoomcamp
• How to Build Production-Ready RAG Systems

How

Tech Stack

• Frontend: Next.js.
• Backend: REST API backend containerized with Docker.
• Database: PostgreSQL for structured data and Qdrant (or pgvector) for vector embeddings
• Cloud Infrastructure: Deployed on AWS using ECS Fargate, with potential usage of Lambda Docker or App Runner.

Dataset

The system integrates personal notes, interview reflections, and structured knowledge

Pipelines

1. Feature Pipeline – collect and structure data
2. Training Pipeline – fine-tune and build embeddings
3. Inference Pipeline – serve and query the model
4. LLM Ops for monitoring and iteration.