Product Engineer (AI-native)

On-site, Dhanmondi 27, Dhaka · Full-time · BDT 50,000–100,000 per month

Who we are

FEMS runs schools in Bangladesh: admissions, student records, fees and accounts, attendance with SMS to parents, exams and report cards, HR and payroll, library, and transport. Schools use it every day, and they tell us quickly when something is wrong.

How we work

We build with AI all day. From August to October 2026, a team of four engineers merged nearly 1,600 pull requests, more than 500 a month on average, most of them written with AI agents (Claude Code and Codex). Work runs in parallel branches and worktrees. Agents follow shared skills and repository instructions, AI reviews pull requests, CI checks every change, and we can spin up a preview environment for any pull request.

Recently shipped:

A student record workspace that puts fees, waivers, library, transport, class routine and applications on one screen
A school calendar with subscribable feeds
Payroll salary heads, tax withholding, and payslips that explain each deduction
Faster attendance reports and departure SMS status for parents
Library workflows and reports
Preview environments for pull requests

The role

You own outcomes, not tickets. Nobody will hand you a spec and check on you every hour.

Talk to schools. Turn a complaint into the real problem, fix it, and close the loop with the customer.
Find what to improve next, in support tickets, error logs, usage data, or by using the product yourself, and ship it without waiting to be asked.
Run several pieces of work at once with AI agents, and make sure what they produce is correct before it merges.
Raise the bar for how we build with AI: better skills, better prompts, better guardrails, less manual work.

You are

A product person who writes code. You care whether a school accountant finishes month-end faster, not only whether the tests pass.
Fluent with current AI coding tools (Claude Code, Codex, Cursor or similar). You're comfortable with agent instructions (CLAUDE.md / AGENTS.md), skills, MCP and parallel agents, and you know where they fail.
Strict about verification. Fast output that breaks a school's fee ledger isn't fast.
Comfortable in a large, real codebase: Laravel (PHP), Vue 3 with TypeScript, MySQL. You don't need to know all of it on day one; you need to learn it fast.
Clear in English and Bangla, written and spoken, with customers and with the team.
Willing to disagree with evidence, then commit and ship. "We've always done it this way" isn't evidence.

Nice to have

You've shipped a product that real customers use, ideally a business system that handles money or records.
Experience with multi-tenant SaaS, payments, or school operations.
Public work you can show us: repos, write-ups, or your own agent setup.

Not a fit if

You'd rather argue about whether to use AI than ship with it.
You need detailed tickets and daily direction to make progress.
You measure your work in hours, not in what changed for customers.

If that's not you, this role isn't a fit. No hard feelings.

How to apply

Apply through the application form. It takes about 30 minutes and ends with short, timed scenario questions. Answer in your own words: we want to see how you think. We don't accept CVs by email.

We review applications every few days. Shortlisted applicants join us at our office for a 60–90 minute session where we build something together using your AI setup.

Apply now

How we work

Recently shipped:

A student record workspace that puts fees, waivers, library, transport, class routine and applications on one screen

A school calendar with subscribable feeds

Payroll salary heads, tax withholding, and payslips that explain each deduction

Faster attendance reports and departure SMS status for parents

Library workflows and reports

Preview environments for pull requests

The role

You own outcomes, not tickets. Nobody will hand you a spec and check on you every hour.

Talk to schools. Turn a complaint into the real problem, fix it, and close the loop with the customer.

Find what to improve next, in support tickets, error logs, usage data, or by using the product yourself, and ship it without waiting to be asked.

Run several pieces of work at once with AI agents, and make sure what they produce is correct before it merges.

Raise the bar for how we build with AI: better skills, better prompts, better guardrails, less manual work.

You are

A product person who writes code. You care whether a school accountant finishes month-end faster, not only whether the tests pass.

Fluent with current AI coding tools (Claude Code, Codex, Cursor or similar). You're comfortable with agent instructions (CLAUDE.md / AGENTS.md), skills, MCP and parallel agents, and you know where they fail.

Strict about verification. Fast output that breaks a school's fee ledger isn't fast.

Comfortable in a large, real codebase: Laravel (PHP), Vue 3 with TypeScript, MySQL. You don't need to know all of it on day one; you need to learn it fast.

Clear in English and Bangla, written and spoken, with customers and with the team.

Willing to disagree with evidence, then commit and ship. "We've always done it this way" isn't evidence.

How to apply

Apply through the application form. It takes about 30 minutes and ends with short, timed scenario questions. Answer in your own words: we want to see how you think. We don't accept CVs by email.

We review applications every few days. Shortlisted applicants join us at our office for a 60–90 minute session where we build something together using your AI setup.