Build SaaS That Actually Works (& Doesn’t Break Next Iteration)
A systems-first approach to building real SaaS products using LLMs as planning tools - not crutches.
Hosted by AI Garage 🇮🇳 with Arvind Girish, Generalist, Nugget (by Zomato)
Date & Time
📅 Wednesday, 14 January
10:00 PM – 11:00 PM IST
📍 Online (Google Meet)
About the Session
Most “AI-built” products break the moment you:
change a requirement,
add a new feature, or
touch the database schema.
This session is about doing the opposite.
In this hands-on, systems-first walkthrough, we’ll show how to build a non-LLM SaaS platform that:
has a stable architecture,
survives iterations,
and uses LLMs only where they actually help (planning, specs, decomposition).
This is not a demo of prompt magic. This is about how real products are built.
What We’ll Build (Lightning Implementation)
We’ll walk through the full lifecycle of a SaaS product, with some prep work done in advance so we can cover everything within 1 hour.
Step-by-step agenda
System Components & Architecture
What components we’ll use
How they plug together using MCP
Where LLMs fit (and where they don’t)
UI/UX & Product Spec Ideation
How to think about screens, flows, and constraints
Turning vague ideas into concrete product specs
Frontend Planning with LLMs
Using LLMs to plan frontend code
Avoiding brittle, auto-generated UI logic
Module-level PRDs
Breaking the product into modules
Writing PRDs that engineers can actually build from
Database Design (Supabase)
Planning tables and relationships
Using LLMs to reason about schema (not blindly generate it)
Backend API Planning
Designing APIs that won’t collapse on iteration
Clear contracts between frontend and backend
Authentication & Finishing the Product
Google authentication
Tying everything together into a working SaaS
Format
⚡ Lightning implementation (with pre-work)
Heavy focus on thinking & structure
Real-world patterns, not toy examples
Live explanations + Q&A
Before the Session (Important)
All participants will receive one of the following before the session:
an instruction doc, or
a custom GPT Gemini NotebookLM setup
This ensures:
we don’t waste time on setup,
everyone follows the same mental model.
After the Session
You’ll receive:
Full recording
Detailed instructions doc
Reusable frameworks you can apply to your own products
Who This Is For
This session is ideal for:
Founders building SaaS products
Product managers & strategists
Engineers tired of brittle AI demos
Builders who care about systems that last
Not ideal if you’re only looking for:
“build an app in 10 minutes” demos
pure prompt engineering sessions
About AI Garage
AI Garage is a builder-led community helping operators, founders, and teams build real AI systems - not just talk about them.
We focus on:
systems thinking,
production-ready architectures,
and learning by building.
Seats & Access
This is a focused, late-night builder session.
Seats may be capped to keep discussion high-quality.
👉 Register to get access to the pre-work.