Software Engineer, AI-Native

About PyjamaHR -

PyjamaHR is an applicant tracking system used by 600+ paying companies to run their hiring. Around the ATS it runs an AI voice interviewer that has conducted more than 65,000 live interviews since 2024, over 3,000 a month, a candidate database of tens of millions of Indian profiles now rolling out, and integrations with LinkedIn, Monster, ZipRecruiter and other job boards. The engineering team is deliberately small: a lead engineer, the founder, and this hire.

How engineering works here -

This is an AI-native engineering team in practice, not as a slogan. Every engineer works inside Claude Code or Codex all day. A typical unit of work starts with a live problem (a support ticket, a failed interview, a recruiter’s screenshot), moves to a root cause proven with production evidence (logs, database, email provider), and ends as a small, well-described pull request. Pull requests are reviewed by several AI models and then by a human who re-derives the reasoning. Descriptions read like incident reports: what happened, why, what changed, how it was verified. There is no ticket queue between you and production. The bar is speed with verification, not speed alone.

What you will work on -

Real examples from the last few weeks, not a wish list:

• Production engineering on the ATS (Python, Django, Celery, Postgres, React): asynchronous task chains that failed silently, email queues that dropped mail, resume parsing that lost files on large batches, LinkedIn profile URLs extracted wrongly from PDFs, job-board feeds with years-old defects.

• The AI interviewer (Python LiveKit agent, Next.js, AWS Transcribe and Polly): speech-to-text reliability, interview-ending logic, caption delivery, cost per interview.

• New surfaces moving from prototype to product: the candidate database (Next.js, OpenSearch), redesigned career pages, an internal leads app. Hardening means tests, feature flags, error states and monitoring.

• The billing rebuild: Razorpay subscriptions and UPI Autopay, job-credit metering replacing the current plan model, migration of thousands of legacy and lifetime-deal accounts, with old and new billing running side by side.

• Infrastructure: moving off a legacy AWS footprint to leaner infrastructure, database retention, cost.

You own tickets end to end: reproduce, fix, test, ship, watch it in production.

Who this is for -

• 0–5 years of professional experience. Exceptional freshers are welcome and compete on equal terms.

• Python first. You have shipped Python that real users depended on (Django preferred; Celery and Postgres a plus). Comfortable in React; TypeScript is a plus.

• Genuinely AI-native. You work inside an agentic coding tool every day, and you have built something with AI that made it do real multi-step work: an agent, a pipeline, an automation that replaced human hours, a system that runs without you. Using AI to write emails, build a portfolio site or fix bugs in a codebase is table stakes here, not an answer. The application asks for the specific thing you built. That answer decides whether we talk.

• You verify. You know when a model is confidently wrong and you have habits for catching it: tests, reading the diff, checking production data before claiming a fix.

• Clear communication. You can explain a bug in five sentences, write a pull request description someone else can act on, and disagree with a senior engineer with reasons.

Working arrangement -

• Bengaluru. In the Koramangala office for the first weeks while you ramp up, hybrid after that.

• Reports to the Lead Engineer; works daily with the founder.

• ₹12–24 LPA fixed, full-time.

Process -

1. Application, about ten minutes, including a short written answer about what you have built with AI.

2. A 15-minute AI interview about your own work and how you think. No trivia, no live coding.

3. A conversation with the founder and the lead engineer.

Decisions within a week of the final conversation.