Building TinyGuide: A Backend Developer's Journey into AI

18 Jun, 2025

Building TinyGuide: A Backend Developer's Journey into AI

It's 3 PM on a rainy Saturday. My toddler is bouncing off the walls, I've exhausted my activity ideas, and naptime is still an hour away. Sound familiar?

After 10+ years of building backends for web and SaaS applications, I'm diving into something completely new: AI engineering. I'm documenting my journey building TinyGuide, an AI-powered parenting assistant, and learning in public.

Who Am I?

I'm a backend engineer with experience in Java, Spring, Go, Kubernetes, and Terraform. I've built distributed systems, managed databases, and deployed services at scale. But here's what I haven't done: built anything with LLMs, vector databases, or AI agents.

I'm also a parent to a toddler who recently asked me "why is the sky up?" at 6 AM. These are the moments that inspired TinyGuide.

What is TinyGuide?

TinyGuide is an AI parenting assistant that helps with:

• Activity suggestions based on age, weather, and energy levels
• Milestone tracking with personalized insights
• Meal planning for picky eaters
• Daily routine optimization (because toddlers thrive on routine)

Think of it as having an experienced parent friend available 24/7 through chat, WhatsApp, or even Alexa.

Why Build This?

Personal Need: Every parent knows the "what should we do today?" panic. Or the "is this behavior normal?" googling at 2 AM. I want an AI that knows my kid's preferences and gives personalized suggestions.

Learning Goals: The AI landscape is moving fast. As a backend engineer, I need to understand:

• How to integrate LLMs effectively
• What RAG (Retrieval Augmented Generation) actually means
• How vector databases work
• Building real-time AI experiences
• The real costs and constraints of AI applications

Portfolio Project: Building a full-stack AI application from scratch is the best way to learn. Plus, it's something I'll actually use, which means I'll push through the hard parts.

The Tech Stack

I'm choosing boring technology where possible and new AI tools where necessary:

Backend (Comfort Zone):

• Go - I know it well, it's fast, great for real-time
• PostgreSQL - Boring, reliable, works
• Redis - Caching and queues

AI Stack (Learning Zone):

• Claude API - Anthropic's LLM for conversations
• Pinecone - Vector database for semantic search
• OpenAI Embeddings - Converting text to vectors

Frontend (Stretch Zone):

• Next.js 14 - Modern React framework
• Tailwind CSS - Utility-first styling
• PWA - Works offline, installable

Infrastructure (Free Tier Focused):

• Fly.io - Backend hosting
• Vercel - Frontend hosting
• Supabase - Managed PostgreSQL

The 12-Week Plan

I'm committing 2 hours per day, 5 days a week (10 hours total) - enough to make real progress without burning out:

• Weeks 1-2: Go backend with auth and basic CRUD
• Weeks 3-4: First AI integration with Claude
• Weeks 5-6: Learning React/Next.js basics
• Weeks 7-8: Building core features
• Weeks 9-10: Implementing RAG with Pinecone
• Weeks 11-12: PWA features and deployment

What Makes This Different?

Most AI tutorials show "build a chatbot in 5 minutes." I'm interested in the real challenges:

• How do you handle rate limits?
• What's the actual cost per user?
• How do you test AI features?
• When should you cache vs. generate?
• How do you handle streaming responses?
• What about offline functionality?

Follow Along

I'll be writing weekly posts covering:

• Technical decisions and trade-offs
• Actual code snippets that work
• Mistakes and how I fixed them
• Cost breakdowns with real numbers
• Performance optimizations

GitHub: github.com/nimbletactician/TinyGuide (Coming soon)
Demo: Will be live at tinyguide.app

Week 1 Goals

Next week I'm starting with the basics:

1. Set up Go backend with Chi router
2. PostgreSQL with user authentication
3. First endpoint that calls Claude
4. Deploy "Hello World" to Fly.io

Questions I Hope to Answer

• Is Go a good choice for AI applications?
• How much does a real AI app cost to run?
• Can you build a good UX with streaming LLM responses?
• Is RAG worth the complexity for small apps?
• How do you prevent prompt injection?
• What's the simplest path to production?

Join the Journey

If you're a backend developer curious about AI, or a parent who codes, or just interested in watching someone learn in public - welcome! This is going to be messy, imperfect, and real.

What's your biggest parenting challenge that AI could help solve? Drop me a note at harikeshkalyanpur@gmail.com or @hakalyanpur.

Next post: "Week 1: First Steps with Go and Claude"

Building TinyGuide in public. Backend developer learning AI, one commit at a time.