NexBot

The Problem I Wanted to Solve

Most small and mid-size businesses in India can't afford enterprise chatbot tools. Intercom, Drift, Tidio — they're either too expensive or too generic. I wanted to build something that could be white-labeled, customized deeply, and deployed affordably.

So I built NexBot.

What I Built

NexBot (nexbotapp.com) is a multi-tenant AI chatbot SaaS platform. Businesses can sign up, train their own bot on their data, and deploy it — without writing a single line of code.

Under the hood:

• 4 Spring Boot microservices — Auth, Bot Engine, RAG, Billing

• RAG pipeline using pgvector + Groq LLM (llama-3.3-70b) — so bots answer from your data, not hallucinated responses

• Stripe billing — subscription management built in from day one

• Full AWS ECS Fargate deployment — containerized, scalable, production-grade

• Single ALB with path-based routing — cost-optimized after I burned through AWS free credits running 4 separate load balancers 😅

• GitHub Actions CI/CD — every push deploys automatically

I did everything myself. Backend, infrastructure, DevOps, database design, billing integration. Solo.

The Hardest Parts

1. AWS costs sneaking up on me I initially ran 4 Application Load Balancers — one per microservice. When my free credits ran out, the bill hit hard. I had to tear everything down, create a new AWS account, and redesign the entire infra around a single ALB with path-based routing. Painful, but I came out with a much cleaner architecture.

2. CORS issues in production Worked perfectly in local. Broke completely on AWS. Spent days debugging before I realized ECS task IPs were changing on every deployment, breaking my API Gateway config. Fixed it with a GlobalCorsFilter with highest-precedence ordering. The kind of bug that makes you question your life choices at 1am.

3. RDS security group configuration Classic — ECS tasks couldn't reach RDS. Security group rules weren't allowing the right inbound traffic. Simple in hindsight, brutal in the moment.

4. Keeping motivation up while building solo No team to keep you accountable. No deadlines. Just you, your laptop, and a half-broken Docker container at midnight. The only thing that kept me going was writing down what I wanted to ship each week and treating it like a real sprint.

What's Working

• Platform is live and stable at nexbotapp.com

• RAG pipeline gives genuinely useful, context-aware responses

• Billing and multi-tenancy are fully functional

• The architecture can scale without a rewrite

What I'm Doing Now

Looking for my first paying clients — targeting Indian SMEs in EdTech, real estate, and healthcare. If you're reading this and you need an AI chatbot that actually understands your business data, reach out.

Also open to feedback from other indie hackers who've gone through the SaaS acquisition grind.

Numbers So Far

• 0 paying customers (honest)

• 1 fully production-deployed SaaS platform

• 4 microservices running on AWS Fargate

• Countless debugging sessions that taught me more than any course ever could

What I'd Tell Myself at the Start

Don't over-engineer the infra before you have users. Ship first, optimize later. I learned this the expensive way.