About

One person, one project at a time, and a bias toward deleting manual work.

Rakib Islam — AI automation, frontend & backend

Rakib Islam — Bangladesh

I’m Rakib Islam. I build AI automation and the frontend and backend systems that carry it — working from Bangladesh, with clients wherever they are. One project at a time, because the last ten percent is where automation either works or quietly breaks.

Every build starts with the boring question: what is this actually costing you per week? If an automation cannot pay for itself in saved hours, I will say so before you sign anything.

I like the problems that sit between departments — the spreadsheet emailed every Monday, the form retyped into a CRM, the report nobody has time to write. That is usually where the cheapest hours are hiding.

2026
Practising since
AI
Automation first
Full
Front & back end
1:1
One client at a time
AI Automation · Business Automation · Frontend · Backend

What I do

Four disciplines. Most projects use two of them, and all of them end in something that runs without you.

Hover a discipline for what you actually receive.

01

AI Automation

Agents and pipelines that take a task off your team entirely — not a chatbot bolted onto a page. Lead triage, document handling, reporting, follow-ups: the work that eats hours and never gets thanked.

  • Workflow audit
  • Agent design
  • Model integration
  • Human hand-off
  • Monitoring
02

Business Automation

The unglamorous plumbing between the tools you already pay for. CRM to invoice to inbox to spreadsheet, connected once and correctly, so nobody is copying rows at midnight.

  • Process mapping
  • API integration
  • Data sync
  • Alerting
  • Handover docs
03

Frontend

Interfaces people actually want to open. React and Next.js, fast on a mid-range phone, accessible by default, and built so the next developer can read it.

  • UI build
  • Design system
  • Responsive layout
  • Performance
  • Accessibility
04

Backend

APIs, databases and jobs that stay up when the traffic is real. Sensible schemas, real error handling, and logs that tell you what happened at 3am.

  • API design
  • Database schema
  • Auth
  • Background jobs
  • Deployment

How it goes

Four steps, and you can stop after any of them.

  1. 01

    Find the leak

    A short call and a look at how the work actually moves today. We count the hours before we touch anything, so there is a number to beat.

  2. 02

    Map it honestly

    Every step written down, including the ones nobody admits to. Some of them should be deleted rather than automated, and I will say which.

  3. 03

    Build the smallest version

    One workflow, running end to end, in front of you early. You judge it against real work rather than a slide.

  4. 04

    Hand it over properly

    Monitoring, a plain-language runbook, and the credentials in your name. It should keep running whether or not I answer the phone.

A quiet identity for a Lisbon ceramics atelier.

Aurelia — A quiet identity for a Lisbon ceramics atelier.

Toolkit

Chosen per project, not per fashion. The list matters far less than whether the thing still runs in a year.

Python
TypeScript
React
Next.js
Node.js
Claude API
OpenAI API
LangChain
n8n
PostgreSQL
Supabase
Redis
Docker
REST & Webhooks
Playwright
Vercel

A short history

A short one, honestly. It starts in 2026 and it is still being written.

2026

Started taking client work

Began building automations for small teams who were losing whole days to work a script could do. Everything since has come from that one observation.

Tell me what you keep doing by hand.