Magica is Hiring SDE-1 (Backend Developer) — Dwarka, Delhi

By Ravinder Singh | The Interns

Most "AI product" job listings you'll come across are vague — a lot of buzzwords, not much substance about what you'd actually build. Magica's listing isn't like that. It's unusually specific about what the job involves and what's genuinely hard about it, which is worth reading properly if you're serious about working on AI-native products rather than just bolting a chatbot onto an existing app.

Quick Details

Company	Magica
Role	SDE-1, Software Engineer (Backend-Heavy)
Experience	1–3 years
Location	Dwarka, New Delhi (On-site)
Compensation	₹6–24 LPA + ESOPs
Apply	Apply Now

What Magica Actually Builds

Magica is building a platform where users hand off work to an AI agent and watch it actually execute — not just respond to a prompt once. The agent runs server-side; the interface is where users describe what they want done, watch sub-agents run in real time, review the output, course-correct mid-task, and approve anything sensitive before it happens.

If you've used tools where AI just generates a static response and stops, this is a different category of product — it's built around ongoing, visible, correctable execution. That distinction is exactly why this role exists.

What You'll Actually Work On

The listing breaks this down into three concrete areas, and it's worth understanding each:

Orchestration UX — handling sub-agent state, streaming partial output as the agent works, exposing memory, and managing interrupts and approvals. This is the core "trust layer" between the user and an AI doing real tasks on their behalf.
Connected services — building the integrations that let users plug Gmail, Drive, Slack, and the rest of their stack directly into Magica.
Channels — keeping the experience consistent across not just the web app, but Magica's presence on WhatsApp, Telegram, and iMessage.

What's Genuinely Hard About This Role

The listing is refreshingly honest about the difficulty here, and it's worth taking seriously rather than skimming past:

You're designing UI for a system that behaves probabilistically and asynchronously. Something as simple as "the agent is working" becomes a real UX problem at every layer of the product.
AI features degrade in subtle, non-obvious ways. Code that works fine in development can quietly break when underlying model behavior shifts — so you're expected to build with monitoring and evals in mind from day one, not bolt them on later.
The codebase moves fast, with regular refactors and migrations. You're expected to stay oriented without waiting on documentation that often doesn't exist yet.

One more thing worth flagging clearly: you'll do most of your actual building through coding agents — the listing states plainly that writing code by hand is considered too slow at this stage. What matters is your ability to direct those agents well and catch when they're wrong. If your workflow is still entirely manual, hand-typed code, this is a meaningfully different way of working than what you're used to.

Skills and Tools You Should Actually Know

The listing doesn't hand you a checklist of frameworks, so read the job itself to infer the stack you should be comfortable with before applying:

Real-time systems — WebSockets or Server-Sent Events, since "streaming partial output" and live sub-agent state are core to the product. If you've never built a feature that pushes updates to a client in real time, spend time on this first.
API integrations at depth — not just calling an API once, but building durable, secure OAuth-based integrations (Gmail, Drive, Slack) that handle token refresh, rate limits, and partial failures gracefully.
Working with LLM-based systems — you don't need to have built a foundation model, but you should understand prompt/response cycles, tool-calling patterns, and why an agent's output can be non-deterministic. If you've built anything using an LLM API, even a small project, mention it.
Observability — logging, monitoring, and evals aren't an afterthought here; they're mentioned explicitly as a day-one expectation. Familiarity with structured logging or basic eval frameworks is a real plus.
Comfort directing AI coding tools — this is stated outright in the JD. If you've used Cursor, Claude Code, Copilot, or similar tools as your primary way of writing code rather than an occasional autocomplete, that's directly relevant experience, not a side skill.

What a Typical Week Might Look Like

The JD doesn't spell this out directly, but reading between the lines of the three focus areas and the "hard parts" section, a realistic week in this role probably involves: picking up an ambiguous ticket (like "make interrupts feel less jarring" or "add Slack as a connected surface"), scoping it yourself without a detailed spec, directing a coding agent through most of the implementation, testing against real probabilistic agent behavior rather than deterministic unit tests, and shipping within days rather than sprint-cycles. If that cadence sounds energizing rather than exhausting, this environment will likely suit you.

Who Magica Is Looking For

No rigid checklist of years-of-experience-with-X here. What they actually want:

A portfolio of real things you've shipped — not coursework, not tutorials, actual shipped work
The ability to take an ambiguous spec, scope it yourself, and ship something usable in days, not weeks
Prior work on AI or agent products is a plus, though not stated as mandatory

Read between the lines here — this favors people who've built and shipped independently (side projects, freelance work, a startup that didn't work out) over people who've only worked inside heavily structured, spec-handed-to-you environments.

How to Apply — This Is Different From Most Listings

There's no form or email here. You apply by sending a LinkedIn connection request to try.magica.com/apply, with a short note that covers:

The role you're applying for ("Software Engineer")
One line on why you're a good fit
Where you found this opening

Important detail from the listing itself: they see your LinkedIn profile along with the connection request — so your profile needs to actually reflect the shipped work they're asking for. Make sure your LinkedIn isn't just a resume copy-paste; if you've built things, they should be visible there before you send this request, not buried in a separate portfolio link they'll never click.

Quick FAQ

Is this role fresher-friendly?
No — it's listed for 1–3 years of experience. If you're a fresh graduate with a strong shipped portfolio, it may still be worth a shot, but go in knowing the stated bar is not entry-level.

Is this remote-friendly?
No, it's on-site in Dwarka, New Delhi. Factor in commute or relocation before applying.

Do I need prior AI/agent experience?
It's listed as "a plus," not a requirement. A strong shipped portfolio matters more than the specific domain of past projects.

Honest Take

This is one of the more substantive AI-product engineering roles I've come across for this list — the JD itself tells you more about the actual day-to-day than most listings twice its length. The compensation band (₹6–24 LPA + ESOPs) is wide, which usually means they're flexible on level depending on how strong your portfolio is, rather than locking everyone into one number.

One thing worth noting honestly: despite being titled "Backend-Heavy," a lot of what's described here — orchestration UX, streaming partial output, channel consistency across web/WhatsApp/Telegram — reads as frontend-and-systems work as much as pure backend. If you're a backend purist expecting zero UI-adjacent work, clarify this directly before applying so there's no mismatch on day one.

The "you'll build mostly through coding agents, not by hand" detail is genuinely worth sitting with. If you enjoy the craft of hand-writing code and find that meaningful, this role's philosophy might not sit well with you. If you're comfortable treating code generation as a tool you direct and verify rather than something you personally type line-by-line, this is a forward-looking way to work and worth the experience.

The on-site requirement in Dwarka is a real constraint — factor in commute or relocation if you're not already based in that part of Delhi. And because the ESOP component is part of the range, it's worth asking directly during the process what stage the company is at and what the ESOP terms (vesting schedule, strike price) actually look like — equity is only worth what the company eventually becomes.

Apply if you have real shipped work to point to and are comfortable with an ambiguous, fast-moving environment. This isn't a role with a settled playbook — that's explicitly the point.

More Opportunities on The Interns

Posted by Ravinder Singh. All listings on The Interns are manually verified before publishing.

Search This Blog

The Interns - Jobs, Internships & Fresher Opportunities in India