As a software engineer, I’ve spent most of my career building backend systems—APIs, distributed systems, and backend services. Over the past year, AI has become an increasingly important part of software development. While I’ve been using tools like ChatGPT and AI-assisted code generation extensively, I realized there’s a big difference between using AI and building AI-powered applications.
This series is my attempt to bridge that gap.
Rather than jumping from one tutorial to another, I decided to learn by building a real project from scratch. For my learning, I m using AI assistant Chatgpt to create my weekly learning plan and Cursor as coding assistant, but the goal is to understand the concepts and engineering decisions rather than generating code.
I’m building AI Design Review Assistant for this project.
As a Senior Software engineer, I spent significant amount of time writing and reviewing software design documents and naturally wanted to choose something which align with what I already do. The idea is simple – upload a software design document and receive a structured feedback similar what you might get from design reviews. Over time, I want it to identify scalability concerns, reliability gaps, observability issues, missing failure modes and questions a reviewer would ask.
What the series will cover:
I’ll start with a simple script and gradually evolve it into a production-style backend application, adding new concepts as I learn them.
Here’s the roadmap I’m following today, with the expectation that it will evolve as the project grows:
- Week 1: Calling an LLM API, prompts, structured outputs
- Week 2: Building a FastAPI backend for design reviews
- Week 3: Embeddings, vector search, and RAG
- Week 4: Document ingestion and knowledge retrieval
- Week 5: Logging, metrics, retries, and observability
- Week 6: Multi-step review workflows and agentic patterns
My hope is that by the end of this series, I’ll not only have a working AI Design Review Assistant but also gain a understanding on what it takes to build a production-ready AI application.
If you’re a backend engineer who’s curious about AI beyond prompts and chatbots, I hope this series gives you a practical place to start.
