Our development team is turning a whole company into a system that improves itself. Every repeated task a colleague does is a candidate for software, and that software is built and checked by AI agents we set up ourselves. You don't inherit someone else's backlog: you decide what gets built and how, you talk directly to the people who use it, and you report to the owner who decides.
The role You take a request from a business user and carry it until it works for them. Which tools, which design, which order: that is your call. You build with AI coding agents as your main way of producing code, check the result, hand it to the users and stay with it until it works in their hands. You plan your own day.
First: web applications that turn engineering data into documents for clients and apartment buyers. You do not need Hebrew; a product manager covers the domain.
Alongside: the developer tooling our team uses every day.
Later: new applications for other departments, starting with business automation. The role grows with you, from building apps with AI agents to designing how our company builds software with them.
How we work
You write the plan before the agent writes code: what the user gets, the first piece that ships within days, and what waits.
You check what the agent built, not just whether tests pass. An agent once wrote our business terms straight into the code; it looked clean and cost 3 developer-days to redo.
When you don't know how long something takes, you build a small test first, then give a date.
You decide without waiting for approval and say how you'll know you were wrong.
Bugs in new work are normal. Tell us about one the day you find it.
What you need to bring
A strong backend, frontend or full-stack background. Any language or stack.
Fluent spoken and written English. Hebrew is not needed.
At least one product you shipped on your own, from a user's request to real users.
Daily work with AI coding agents as your main way of producing code.
Who you are These are Midved's five company values, written as what they look like day to day.
Resilience and self-awareness. You keep ego out of the work, take feedback as a tool to grow, and stay calm under load.
Trust and autonomy. You own a task from end to end. If something slips, you say so early, before anyone asks.
Professionalism. You keep learning and master your tools. You understand why each part of the work exists, and everything you ship carries your quality signature.
Productivity and effectiveness. You define the result, break it into steps and decide what not to do now. You measure yourself in tasks closed, not hours worked.
Force multiplier. The team's success is your success. You share what you know and help others get better.
This role is not for you if
You want detailed tickets and someone to review every line you ship.
You write code mostly by hand, or use AI only for autocomplete.
You can't recall a time an AI agent gave you something wrong that looked right.
You measure your week in hours, not in things that work for users.
How to apply — we read applications that include items 1 to 3
A link we can check: a live product, a public repository or a demo of something you shipped on your own, plus one sentence on a decision you made in it and why. Under NDA? Say so, and you'll walk us through it on screen in the call.
Your monthly salary expectation in USD.
Two or three sentences: the last time an AI agent gave you something wrong that looked right, and how you caught it. Optional, and we watch these first: a 5-minute screen recording of you working with your AI agent on a real task. Unedited is fine.
How we hire Your application (we answer within 2 working days) → a 45-minute call in English with our Head of Development, including a screen share → a paid test day ($170): you plan a real change to one of our systems, then defend the plan live → a conversation with our CEO → references. From application to of
