Full Stack AI

Full-Stack Product Engineer, AILocation — Hyderabad, India (Hybrid)Employment type — Full-timeExperience — ~3–6 years, or a body of shipped work that speaks for itselfCompensation — Competitive with marketABOUT AIRAWATHAirawath is an AI-powered e-invitation and event platform. People describethe celebration they're planning; we generate the invitation design, deliverit over SMS, email, and push, track RSVPs, and run the gift registry aroundit. Elie, our recommendation layer, helps guests find the right gift.We're a growing in-house engineering team working directly with the founders.ABOUT THE ROLEProblems here don't arrive divided into frontend and backend tickets. You'llown features end to end — understand the user problem, design the schema,build the APIs and the interface, integrate the AI layer, deploy it, andmeasure whether it changed anything. You'll own systems rather than a queue,and you'll have the latitude to decide how they get built.RESPONSIBILITIESIn your first year you'd likely own two or three of these outright:• AI invitation generation. The text-to-image pipeline behind every design — prompt construction, model selection, cost and latency per generation, and the eval loop that tells us whether output quality actually moved.• Notification delivery. One invitation fans out to SMS, email, and push across timezones. Deliverability, retries, idempotency, and the telemetry that proves a guest was genuinely reached.• Credits and payments. The credit economy behind AI generation, and the payments layer on top of it. Money code — the highest bar in the codebase.• Elie, gift discovery. Parameter-based and conversational recommendation over a live product catalogue. Retrieval, structured outputs, and the judgment to know when an answer shouldn't come from a model at all.For any of them you'd write the migration, the API, the web surface, the AIlayer, the deployment, and the dashboard that says whether it worked.MINIMUM QUALIFICATIONS• Production React, TypeScript, and Next.js. You've shipped and then maintained real interfaces, not just built them once.• Production Python, including async. Our backend is FastAPI on SQLAlchemy 2.0 async, where the interesting problems are session and transaction semantics rather than syntax.• PostgreSQL judgment. Schema design, relationships, indexing, and the ability to read a query plan. We hand-write SQL migrations and apply them to a live production database.• You've shipped an LLM-powered feature that real people used. Not a demo — you can speak to what it cost per call, where it failed, and how you knew it was working.• You've deployed and debugged your own code in production. AWS (EC2, RDS, S3), Docker, and CI. When something breaks at 11pm, you're the one reading the logs.PREFERRED QUALIFICATIONS• AI agent or multi-agent workflows; RAG, embeddings, vector search, and retrieval evaluation.• Evaluation frameworks for LLM accuracy, reliability, latency, and cost.• Flutter or native mobile development — we have a mobile app, and you'd occasionally touch it.• Streaming and real-time systems: WebSockets, server-sent events, voice.• Redis, queues, event-driven architectures, AWS Lambda or ECS.• Observability, logging, monitoring, and production analytics.• Early-stage startup experience, or open-source and independently launched work.HOW WE WORK• Own the outcome. You're accountable for whether the feature worked, not whether it shipped. Everything you own gets a metric you chose.• Short decision paths, real review discipline. You don't need a committee to pick an approach. But anything touching money, authentication, or migrations gets read line by line — yours and ours.• AI by default. Cursor, Claude Code, Codex — whatever gets you there. We care that you know where the tools are unreliable, not that you use them.• Keep it practical. The simplest reliable thing that creates impact. We'd rather delete code than add a layer.Airawath is an equal opportunity employer. We consider all qualifiedapplicants without regard to race, religion, colour, sex, sexual orientation,gender identity, age, national origin, disability, marital status, or anyother characteristic protected by law.