AI-Native Engineer (Full Stack)
Own HeyHomie’s AI engineering layer from day one — and grow into AI Lead.
Location
Bangalore (in-office)
Work model
On-site
Type
Full-time
Experience
3–6 years
Compensation
Senior IC + ESOPs
Reports to
Founder & CEO
Trajectory
AI Lead in 12–18 months
WhatsApp isn’t where India chats anymore. It’s where India transacts.
The layer being built on top of it is no longer features. It is intelligence.
The next 24 months decide who owns the AI layer in Indian conversational commerce — whether sellers across categories run their entire businesses through agents that understand intent, context, language, inventory and price. Whether AI becomes the operating system for India’s small-business economy, or just another bolt-on no one ends up using.
We are placing our bet. And we are looking for the person who builds the bet with us.
“This is not a “join an AI team” role. This is the role that builds the AI team.”
You own the AI engineering layer of HeyHomie — end to end, from day one.
You are an AI-native full-stack engineer who reports directly to the founder and owns the AI engineering layer of HeyHomie from day one.
You write the production code. You design the architecture. You make the calls on stack, models, retrieval, agents, evaluation and infrastructure. You ship to production from week one — no product committee or stage gate between you and impact.
The trajectory is explicit, not implied. As we scale, you grow into AI Lead: you own the AI engineering function, you hire around you, and you set the technical direction for what HeyHomie becomes over the next three to five years.
This is a founding-engineer-grade hire — with the responsibility, the equity and the autonomy that goes with it.
You own the AI layer. End to end.
The conversational AI stack
The agents that run on WhatsApp and Instagram for sellers across 22+ categories. Intent classification, context retention across sessions, multi-turn conversations that actually convert, multilingual handling across Hindi, English, Hinglish and regional languages — and graceful fallback when the model is wrong.
The retrieval & grounding layer
Every seller’s agent answers as that seller would, not as a generic model would. You own the vector strategy, the embeddings, the hybrid retrieval, the evaluation harness and the freshness pipeline that keeps the agent in sync with catalogue, pricing, inventory, policies and recent conversations.
The agent orchestration layer
Sellers configure flows without writing code. You build the multi-agent handoffs (sales → support → upsell), the tool-calling for live inventory and order lookups, the human-in-the-loop escalation paths, and the guardrails that keep all of it safe at scale.
The full-stack platform around the AI
APIs, webhooks, queues, integrations with Meta’s WhatsApp Business and Instagram Messaging APIs, payments, seller-facing dashboards and config UIs. Backend in Node.js and TypeScript. Frontend wherever a seller needs to see, configure or measure something the AI is doing.
Evaluation & observability
We do not ship AI features without knowing whether they work. Quality metrics, hallucination tracking, latency and cost dashboards, escalation rates, conversion attribution — the dashboards and the discipline that go with them.
Infrastructure & economics
Model selection, prompt economics, caching strategy, inference cost, latency budgets. Not “use the latest model for everything” — the role that figures out how to deliver enterprise-grade AI to a seller paying a few thousand rupees a month.
Too senior for a week-in-the-life. Here’s your first quarter.
We’re not writing a week-in-the-life — the role is too senior and the work too cyclical. Here is what your first 90 days actually look like.
Weeks 1–4 · Understand
You sit with the founder, the engineering team and three to five real sellers. You audit what is already running on the AI side and find the gap between what we have and what the market is about to demand. By the end of week four you have written a one-page architecture note on where the AI layer goes from here. Not a slide deck. A document.
Weeks 5–8 · Ship something that matters
Not a demo. Not a proof of concept. A real system a real seller turns on, runs for two weeks and reports back on. An AI sales agent for a single category, a retrieval-augmented support flow, or something we have not thought of yet. You decide what wins fastest, build it, ship it, measure it, and defend the decisions in front of the team and the seller.
Weeks 9–12 · Set the architecture
Based on what worked and what did not, you write the detailed second draft, start identifying the next two engineers we need, and begin the AI Lead conversation in earnest. By quarter’s end you can point to one production system, one architecture document and one clear thesis on what HeyHomie’s AI layer looks like 18 months out. That is the bar.
A builder, not an explorer.
- A builder, not an explorer. You have shipped real systems to production where scale matters and downtime costs revenue — and can point to specific systems, decisions and trade-offs you owned. Not projects you “contributed to.”
- AI-native in the production sense, not the tutorial sense. You have built agents, RAG systems, evaluation pipelines and prompt economics that run in production. You have a point of view on LangChain vs LangGraph vs LlamaIndex vs writing it yourself — with evidence behind it.
- Full stack with a backend spine. Strongest in backend — Node.js, TypeScript, NestJS or comparable: APIs, data modelling, queues, webhooks, integration plumbing. And you ship the frontend that exposes it (React, Next.js, Vue) without needing two other people in the loop.
- Comfortable with distributed systems. Queues, idempotency, retries, eventual consistency, observability, cost. Bonus if you have worked on high-throughput billing, payments, messaging or commerce infrastructure.
- Modern AI tooling in your fingertips. Vector databases, embeddings, evaluation tools, MCP servers, agent frameworks. You know the difference between what is hyped and what is useful.
- You write code that is read more than it is written. You document what matters, test what breaks, and think about the engineer who works on this in six months — even if that engineer is you.
- A founder at heart. You read the user and the market, hold business and technical context in the same sentence, and can take the call without a PM. A sharp, kind communicator who does not waste anyone’s time.
Not a “maybe one day” promise. The structural design of the role.
Months 0–6 · The Engineer
You ship. You learn the platform, the sellers, the data. You build the trust of the team and the founder, and set the bar for what AI engineering looks like here.
Months 6–12 · The Architect
As the AI layer compounds, you define the team around you — hiring rubrics, interviewing the next two engineers, splitting work across people who report to you. You own architecture decisions across the full AI stack.
Months 12–18 · AI Lead
You own the function: the roadmap, the stack, the hiring, the technical direction. You report to the founder, with a seat in the room where the company gets shaped.
The path is owned by you. The faster you build, the faster you lead.
What this role pays you that a bigger company can’t.
We pay competitively for the senior IC bracket, with meaningful ESOPs — discussed openly at offer stage, with no games. But if compensation is the first and loudest question, larger companies will outbid us. They have to; they have less to offer otherwise. Here is what they can’t match:
The compounding capital of being early
You’ll be three years ahead of the market on AI-native conversational commerce by the time it becomes mainstream. That is a different kind of compounding than a salary number.
Direct founder access, every day
Every architectural call, every product call, every business call — you are in the room. You do not schedule a meeting to influence the direction of the company.
AI leadership at a Meta Technology Partner
A rare seat in a young category — owning the bet, not implementing someone else’s AI strategy. The thesis, the architecture, the team. Yours.
A platform that already has scale
Sellers across 22+ categories. Live transactions. Real conversations. You build for a market that exists, not one you have to manufacture.
“If you read all of that and the loudest thing in your head is still the cash number, this is not the role. We will not be hurt if you decide that. We will be hurt if you join and decide it after.”
You should apply if
- You have shipped AI systems to production — not to hackathons, demos or weekend projects.
- You can show the work: public repos, deployed apps, written architecture notes, talks. Anything that proves the work and the thinking are yours.
- You have built backend systems at meaningful scale in TypeScript, Node.js or comparable, and can talk about the specific bottlenecks you hit and how you solved them.
- You already have an opinion on the WhatsApp and Instagram commerce problem space — before reading this JD.
- You want the AI Lead trajectory and are willing to earn it through the work, not the title.
- You want to be in the room where the company gets built, not the room where someone else’s decisions get implemented.
Do not apply if
- Your CV is mostly responsibilities, not outcomes. We do not care what you were “involved in” — we care what shipped because of you.
- You have explored AI through tutorials, courses or side projects but have never owned a production AI system.
- You need a PM, a designer and a tech lead to ship a single line of code to a user.
- You want to optimise the next 18 months for cash. Optimise for compounding learning here, for cash elsewhere — both are valid, only one fits this role.
- You believe AI is a layer you bolt on, not a layer you architect from the ground up.
- You believe “founding engineer” is a title rather than a way of operating.
How to apply
We read every application.
- 01Hit “Apply for this role” and complete the form. Include your CV.
- 02Add links to two systems you have built that you are most proud of — public repos, deployed apps, write-ups, talks. Whatever shows the work and the thinking best.
- 03Write a 300-word note to the prompt: if you had to build a WhatsApp-native AI agent that helps a homepreneur sell across 22 categories, where would you start, what would you deliberately not optimise for on day one, and what is the one thing you would obsess over? Tell us how you think, not what you already know.
- 04The 300-word note is the most important part — we read it before the CV. Treat it accordingly.
We are placing our bet. We are looking for the person who builds it with us.
