AI Product Engineer
You will build the part of Frep that thinks: the recommendation engine that plans a household’s week, the AI pipelines that enrich our recipes and create their food imagery, and the app and APIs that put all of it in front of families.
Location: Aundh, Pune (in office, five days a week)
Employment: Full-time
Experience: 3–5 years, hands-on
Reports to: Founders
Yes, we're an AI company. No, AI didn't write our recipes.
Every one of Frep's 8,000+ recipes comes from a real kitchen. We have curated them over time from real people: the kind of recipes people defend with their lives.
We're preserving India's culinary heritage, one family recipe at a time. Most of us take maa ke haath ka khaana for granted, until we move away from home. Frep lets you save those recipes and pass them on, wherever you are. Nobody cooks it quite like your mom, and we won't pretend otherwise. But we can get close enough to bring the memories back.
So where's the AI? In the planning. Most of us lose 30 to 40 minutes a day staring into the fridge, working out aaj khaane mein kya banaye. Frep learns what a household likes and can't eat, tracks the pantry, and plans the week around it. It shows you what you have and what you can make, and pushes whatever's missing straight onto your grocery list. Short on time? There's a whole section of recipes ready in 30 minutes or less.
The result: 30+ minutes back every day and 72% less food wastage, which means less money in the bin too.
That planning brain is what you will build.
The Role
You will build the part of Frep that thinks: the recommendation engine that plans a household’s week, the AI pipelines that enrich our recipes and create their food imagery, and the app and APIs that put all of it in front of families.
This is a hands-on role. You will write the prompt, the pipeline, the API, and the React Native screen, and then measure whether the plan is one a real family would cook. If the engine sends pyaaz-lehsun to a Jain household, karela to the one person at the table who hates it, or peanut chutney to a kid with a nut allergy, that’s yours to fix. Two of those are funny. One isn’t. That’s the job.
Your Scope of Work
Your responsibilities will include, but are not limited to:
- Building AI-powered features end to end, from prompt and pipeline design to the API and the React Native screen.
- Improving our recommendation and personalization engine, so every plan is one a real household would cook.
- Building and maintaining AI pipelines for recipe enrichment and food imagery.
- Designing evaluations for AI output, so quality is measured rather than assumed.
- Keeping model calls fast, reliable, and cost-efficient in production, token by token.
- Developing Node.js APIs and backend services on PostgreSQL, vector search, and Redis.
- Diagnosing production issues across the app, the backend, and the AI layer.
- Optimizing performance across the app: no laggy screens, seamless on any phone, and smart about patchy mobile networks.
Life at Frep
Hum teen log hain: two co-founders and one AI/ML engineer, in a five-seater co-working office in Aundh, Pune. Three seats are taken. The other two are for the people we’re hiring right now.Everyone sits at one table. Lunch gets shared in the cafeteria, and food opinions get defended loudly. Rahul, one of our co-founders, will defend biryani to the death. He once burnt his mom’s special mutton akhni biryani to a crisp at the bottom, then served it to the family as “smoked mutton akhni biryani”. We call that product thinking.
- Frep is a Singapore-registered company, built in Aundh, with co-founders across Pune and Singapore.
- We’re in beta testing now and gearing up for a full launch soon. Join now and you’re here for launch day.
- Here, not using AI is the red flag. We pay for Claude Code, and we expect you to be properly good with it.
- We’re bootstrapped today and getting ready to raise funding to take Frep across India and beyond. Every rupee is spent carefully, and every hire changes the company. Prove yourself here and you’re part of the founding team, with all the room to grow that comes with it.
You’ll Fit Right In If
- You explain things clearly, in writing and out loud. We build with Claude Code, so plain English is effectively our programming language. A vague instruction produces vague code.
- You have shipped AI to real users and learned exactly where it breaks.
- You watch token usage the way your mum watches the ghee.
- You care whether a feature is useful, not only whether it works.
- You would rather own a problem end to end than wait for a ticket.
- Your engineering fundamentals are strong. AI makes a good engineer faster. It does not make anyone a good engineer.
Strong English communication is non-negotiable.
What You Will Need
- 3–5 years of hands-on experience building and shipping production software.
- At least one LLM-powered feature that you built and that reached real users.
- Practical experience with LLM APIs, including prompt design, structured outputs, tool calling, and keeping token costs in check.
- Strong TypeScript skills.
- Experience with Node.js, REST APIs, and background jobs.
- Working knowledge of SQL and production databases.
- React Native experience, or strong React experience with the ability to transition quickly.
- Daily use of AI development tools such as Claude Code, Cursor, or Copilot.
Good to Have
Experience with any of the following is useful, but not required:
- Retrieval-augmented generation (RAG), embeddings, and vector search.
- LLM evaluation and prompt versioning.
- Recommendation systems, ranking, or search relevance.
- Generative image pipelines.
- Python for data or ML work.
- Expo and EAS.
- PostgreSQL or Supabase.
- App Store and Play Store releases.
- Consumer apps in food, health, lifestyle, or other high-frequency categories.
Our Stack
Mobile: React Native, Expo, TypeScript
Backend: Node.js, TypeScript
Data: PostgreSQL, vector search, Redis
AI: a recommendation engine for meal planning; LLM pipelines for recipe enrichment and food imagery
Internal tools: ReactInfrastructure: Cloudflare R2, GitHub Actions
How we build: Claude Code, Cursor
How We Hire
- Apply on Naukri, or through our careers page.
- A 30-minute call with a founder, about something you have built.
- A take-home assignment: about 3–4 hours of work, with a few days to do it, so a weekend works. Use any AI tools you like. We’d be worried if you didn’t.
- A final video call where we go through your assignment together. It’s a two-way street: you see how we think as much as we see how you work.
- If it’s a fit on both sides, we make you an offer.
You will hear from us within a week at every step.
Location
This is a full-time, in-office position in Aundh, Pune.
The team works from the office five days a week.
If you are based outside Pune, you should be willing to relocate.
