Applied AI Engineer

Who we are

Supernova is building AI that helps the world communicate. Backed by KAE Capital, Lumikai, and founders from Swiggy, 1MG, Uni, and more, we’re building a communication AI platform that powers two products: Supernova, a communication coach with 10M+ Play Store downloads and a 4.7 rating, and Superflow, a daily copilot past 1M downloads. Both are top-10 apps on the Play Store and the App Store.

Traction

• India’s largest consumer AI startup — profitable and growing fast, with a large paying user base, built and run by a five-person engineering team.
• 15M ARR and scaling fast.
• Superflow, our speech-to-text product, launched two months ago and is already at a 1 Cr/month run rate.
• Millions of users, half a million new transcriptions a day.

Our mission

• Help people communicate clearly and confidently, across languages and everyday situations.
• Give everyone an AI partner that helps them express their ideas, understand others, and handle communication throughout their day.
• Bring communication coaching and real-time assistance together on one shared AI platform, starting with English and building for a global audience.

Why this role exists

Superflow takes a user speaking Hindi, Tamil or Telugu and returns polished English, in the tone they choose, on an Android and iOS keyboard. Getting that right is not a model procurement problem. It is a product problem made of a hundred small judgements: what the prompt asks for, what the model is allowed to correct, what happens when a second voice appears in the recording, how much latency a user will tolerate before they give up and type.

You will own that layer across the shared AI platform behind both Supernova and Superflow. You are the person who treats the LLM as a component in a product and makes it behave.

What you will work on

• Design and iterate the prompts and the LLM pipeline behind Superflow, our speech-to-text product — tone control (casual, formal), latency budgets, and output quality across Indian languages — on the shared platform that also powers Supernova.
• Build the evals that tell us whether a change actually helped. Ship behind experiments, read the numbers, keep what wins.
• Solve real quality problems end to end. Example: about 3% of recordings contain a second speaker — background noise, someone on a call nearby. We want that at 0.1%. The fix may be in the prompt, in the app’s handling of volume levels, in the ASR layer, or all three.
• Run open and hosted models in production — we currently serve on Gemma 4, and we care a lot about what the smallest model that does the job is.
• Build product, not just pipelines. Everyone here picks up whatever the work needs. Nobody on the team had written native Android or iOS before we shipped a third-party keyboard; all of us have contributed to that code now.

How we work with AI

AI agents write most of our code. Engineers direct that work and own the result. Generating code is straightforward; the harder work is deciding what to build, choosing an architecture, judging whether the implementation is sound, and knowing when it is ready for users. You must understand the systems you work on even when an agent performs the implementation.

For any production change, you should be able to explain its architecture, data flow, failure modes, and important design choices — why you chose one approach over another, what happens under unusual or malicious input, and where the system is likely to fail. Before opening a pull request, inspect the implementation, test its behavior, remove unnecessary complexity, and confirm the change fits the surrounding system. Work produced by an agent is still your work — don’t pass unchecked agent output to a reviewer and make them find the mistakes. Pull requests should be small, focused, and ready to merge.

Experiments and disposable prototypes have a different standard: move fast to find the right product, then sustain that pace through the work required to make it excellent. AI makes the first 90% of a feature easy; much of the quality lives in the rest — edge cases, copy, loading and failure states, performance, and coherence. Shipping starts the feedback loop, and we want engineers who keep the urgency after the demo works and refuse to ship slop.

What we are looking for

• 3-7 years of engineering experience, with at least one AI product you took to real users and kept working after launch.
• Strong prompt engineering — not prompt collecting. You can explain why a prompt failed and what you changed.
• Comfort with evaluation: you build the dataset and the harness before you trust a result.
• Solid software engineering fundamentals. APIs, latency, cost, failure modes. You should be able to read and write production code in the stack you are given.
• Can build a working model of an unfamiliar system — its architecture, data flow, failure modes, and constraints — and explain and defend the major implementation choices behind any production work you own.
• Reviews agent-written code with the same rigor as hand-written code, takes full responsibility for its quality, and opens small, focused pull requests.
• Taste. Agents produce a lot of slop and the hard part is refusing it. We want someone who reviews what the agent wrote, keeps the codebase clean, and does not push that review cost onto their teammates.
• A generalist instinct. Titles here are thin; if the problem needs a native keyboard extension, you learn to write one.

Nice to have

• Speech-to-text, ASR or audio pipeline experience.
• Experience serving open-weight models (Gemma, Qwen, Whisper family) with hard latency budgets.
• Consumer-scale India products, or products built for users who are not comfortable in English.

Qualities we love

• Strong engineering judgment. You can distinguish code that merely works from code that belongs in a production system.
• Moves fast and finishes well. You test ideas quickly, then sustain that pace through the work required to make the product excellent.
• Cares about craft. You notice rough edges, exercise taste, and take pride in shipping a polished experience.
• Relentless about improvement. You study real usage, act on feedback, and keep raising the quality after launch.
• Reduces the load on the team. Your work arrives considered, checked, and ready to use. Teammates rarely need to untangle it or discover the basic failure cases for you.
• Ambitious and eager to grow. You want larger problems, higher standards, and more responsibility.
• Contributes beyond assigned tasks. You notice problems and opportunities, propose improvements, and help raise the capability of the team.
• Empathetic. You care about what users actually experience, including details that are difficult to measure.

What we offer

• 40-80 LPA cash plus generous ESOP, and relocation support.
• A five-person engineering team carrying the whole business — your work is visible in the P&L within weeks.
• Cutting-edge work at India’s largest consumer AI startup, on products millions of people use every day.
• Flexible timing and work-from-home days.
• An inspiring office in HSR, Bengaluru, and direct mentorship from experienced founders.