Full Stack Developer, Rails and React

Before you apply

We're direct about how we work. Read this section first.

We value engineers who know what they don't know, and say so. If you're honest about your gaps, we'll invest in closing them. If you hide weaknesses or oversell your experience, you won't last here.

We expect you to use AI coding tools. Everyone here does, and pretending otherwise in 2026 would be theatre. What we care about is what happens next. Roughly two in five lines of committed code across the industry are now AI-written, while fewer than half of developers say they always verify that code before committing it, and in a controlled security study 45 percent of AI-generated samples introduced a vulnerability from the OWASP top ten. The failure mode is specific, and it is not sloppiness. AI code arrives with clean syntax and consistent naming, so the signals that used to make bad code obvious are gone. What hides underneath is a misunderstood requirement or an edge case nobody considered. Reading it for style tells you nothing. You have to work out what the change was supposed to do and then check whether it does that.

So the skill we hire for is not typing and it is not prompting. It is knowing when you are done. You can delegate the work. You cannot delegate the accountability.

We are also not going to tell you that AI makes everyone ten times faster. The one properly controlled study on this found experienced developers were 19 percent slower with AI while believing they had been faster, and the follow-up could not be completed because developers refused to work without the tools. The productivity question is genuinely open. The verification cost is the part that has actually been measured.

We move fast. Growth happens outside comfort zones. If this sounds harsh, we're probably not a fit. If it sounds like the environment you've been looking for, keep reading.

The role

Allerin builds AI and IoT products: edge to cloud video intelligence, computer vision systems, and data platforms. We also partner with select clients to solve their hardest technical problems. You'll do both.

You'll work across a Rails API backend and a React admin frontend, on a production system with roughly 160 models, 231 controllers, and five separate API surfaces covering dashboard, mobile, valet, kiosk, and AI/ML integrations. This is not greenfield. You are joining a system that already has users, and the work is to extend it without breaking what runs.

• Location: Navi Mumbai
• Team: Engineering
• Works alongside: the Backend Team Lead, who owns architecture, and the test automation engineer, who owns the automated suites

What you'll own

Features, end to end

• Take a feature from schema and migration through the API to the React screen a person actually uses, and stay with it after it ships.
• Design endpoints with clean contracts, sensible pagination, consistent errors, and no surprises for the clients that consume them.
• Write the tests that would catch your own mistake, not the tests that pass.

The surfaces where AI meets a person

• Build the screens where a model's output reaches a human: review and override queues, confidence shown honestly, audit trails of who decided what.
• Treat a prediction as a probability rather than a fact. A vision model at 97 percent confidence is still wrong often enough that the interface has to say so.
• Make the numbers visible. If a client is told false positives fell 85 percent, something you built is where that number comes from.

Verification

• Know what you ran, what it proved, and what it did not. "The tests pass" is the beginning of that sentence, not the end of it.
• Review AI-written changes, including your own, for intent rather than style.
• Break your own work before someone else does, and say plainly when it is not ready.

A codebase you did not write

• Read code, git history, and other people's decisions before changing them.
• Ask why something is the way it is before deciding it is wrong. Occasionally it is wrong, and then you fix it properly.

What we expect you can do

We don't count years. We look for evidence, because someone can do a thing for five years and have nothing to show for it. Bring work we can look at and talk about, and be ready for us to go deep on it.

• Build and ship a full feature in a server framework and a modern JavaScript frontend. Rails and React are our stack, and Rails experience is a real advantage, but a strong engineer from Django, Laravel, Node, or Spring who wants to work in Rails is not disqualified.
• Model data properly in a relational database, and read a query plan when something is slow.
• Debug something you did not write, from a symptom, without guessing.
• Use AI tools well, and describe where you stop trusting them.
• Write clearly. Half this job is explaining a change to someone who was not in your head when you made it.

No degree requirement. We will look at what you have built.

Nice to have

• Rails at production scale, including background jobs and API versioning
• Work on a product where a machine learning model's output reached a user
• Performance work with real numbers attached, before and after
• Anything you have written or open sourced that shows how you think

This role is NOT for you if

• You want tickets specified down to the last detail before you start.
• You ship what an assistant produced without being able to say why it is correct.
• "The tests passed" is your complete answer to whether something works.
• You want to avoid the legacy parts of the system and only build new things.
• You need a title before you are willing to own an outcome.

What we offer

• Real ownership of features that reach production and get measured
• A senior engineering team that reviews code properly and expects the same of you
• Investment in your growth if you're honest about where you need it
• Problems that are genuinely hard, on systems that genuinely run

How we hire

Apply through the form. The questions are built to surface how you think, not how well you can prompt a language model. A human reads every application, and answers that read as model-generated are tested in depth in the interview, where they do not survive. Everything you write will be discussed in depth if we talk.

If we do talk, one round is a code review. We hand you a change an AI tool wrote against a small Rails application, with a real defect in it, and ask you two questions. Is this safe to merge, and how do you know. Bring whatever tools you normally use. We are watching how you decide you are finished, which is the part of the job that got harder.