Junior AI Engineer - Backend Python

JuiceLabs AI; Bengaluru; On-site; INR 1,200,000 - 1,800,000; Full time

Job summary

JuiceLabs AI in Bengaluru seeks a disciplined software engineer to help shape zero-to-one AI products. You’ll design systems, write specs, review code and harnesses, and bring ideas into real, shipped tools. 0–3 years of work, internships or open source count.

You’ll work with Python (FastAPI), Postgres, and cloud stacks, and you’ll need clear written communication. We value ownership, fast cycles, and practical demos.

Qualifications

• Proven ability to design systems with clear data models and failure modes.
• Ability to write specs/PRDs as primary source of truth for development.
• Experience reviewing complex code and agent outputs.
• Familiarity with deploying AI-powered tools and harnesses.

Responsibilities

• Design the system boundaries, data models, and failure modes with the CTO.
• Write the spec/PRD to guide implementation.
• Review all PRs, agent outputs, and runtime behavior.
• Bring your AI harness ideas (Claude/agent tooling) and advocate reliability.

Skills

System design

PRD writing

Code review

AI tooling

Communication

Startup mindset

Tools

Docker

CI/CD

Python (FastAPI)

PostgreSQL

Cloud (GCP/AWS)

Job description

Where creative engineering meets applied AI.

We build vertical, AI-native tools that unlock fresh insights and creative superpowers for clients in advertising, ecommerce, and beyond. We're a nimble team of founders and builders obsessed with using AI to solve real problems in ways that didn't exist eighteen months ago.

The Opportunity

Writing code is now the cheapest part of building software. Claude Code and Codex do the typing. What they can't do is decide what's worth building, hold a system in their head, or notice that the confident thing they just produced falls over at 10x traffic.

That layer — above the code — is the job. It used to take five years of ticket work to reach. It doesn't any more.

You’ll work directly with our Cofounder/CTO on zero-to-one products, with someone senior in the loop while you build the judgment to run things alone.

What we screen on: show us what you've built

We mean something you chose, scoped, and shipped because you wanted it to exist. A tool that killed your own annoying problem. A weird agent harness. A scraper, a game, a bot. A thousand users, or exactly one user who was you.

We care that it's real, that it runs, and that you can defend every decision in it. Two projects you're proud of beat a spotless CV with nothing behind it.

What you’ll do:

• Design the system — boundaries, data models, failure modes. Alongside the CTO at first, then on your own.
• Write the spec. PRDs are the new source code. A vague spec produces confidently wrong software, fast.
• Review everything — human PRs, agent PRs, your own 2am output. The line between "it runs" and "it's correct."
• Bring your harness. CLAUDE.md, subagents, eval loops, whatever makes agents reliable rather than impressive. You may be ahead of us here. We want to steal your ideas.

What we're looking for:

• Interesting projects. The strongest signal we have.
• High agency. You see the gap and close it, without chasing.
• Systems thinking. You may not have been paged at 3am yet. You should still know when an elegant-looking solution is quietly quadratic. LeetCode ninja not required; explaining why this won't blow up at 10x, very much required.
• You write well. Writing is thinking, and it's most of how you'll direct both humans and models.
• You've deployed something. Put it on the internet, watched it break, fixed it. Docker, CI/CD, some cloud. We're Python (FastAPI), Postgres, GCP/AWS, but how you think matters more than what you've memorised.
• You actually drive AI tools. Not "I've tried Copilot." You have opinions on where agents lie and how you verify what comes back.
• Startup energy. Multiple hats, fast cycles, and the judgment to balance shipping now against surviving real users.

Experience:

0–3 years. Internships, contract work, open source, hackathons, and serious side projects all count. New grads and self-taught engineers welcome.