Forward Deployed Engineer

Spinny is India's largest full-stack used-car platform. We sell a car every few minutes. Behind every sale sits a chain of conversations — discovery calls, test drive scheduling, post-booking handholding, financing nudges, delivery coordination. Today, much of this runs through inside sales and call center teams handling tens of thousands of calls a day.

We're rebuilding this entire layer with agentic AI. Our in-house product already handles a meaningful share of customer conversations end-to-end — voice, chat, scheduling, follow-ups. We're now scaling it across query types, languages, and workflows that have never been automated in Indian auto-retail before.

This is not a research role. This is not a let's pilot an LLM role. This is a build-deploy-measure-iterate role where the scoreboard is test drives booked, calls deflected, and conversion lifted.

Experience Required: 1-3 Years

What you'll actually do

Sit with inside sales leaders, listen to 50 calls, find the one workflow where a bot can outperform a human — then build it. Design and ship voice and chat agents on production traffic. Wire them into our CRM, telephony, scheduling, and inspection systems. Write the prompts, tune the retrieval, debug the hallucinations, fix the latency, watch the dashboards, and rip out what doesn't work. Own the loop from I have a hunch to this is now handling 10,000 calls a week.

You'll work directly with the CEO's office on problems that show up in board reviews.

What we're looking for

Curiosity over credentials. We care more about what you've built on weekends than where you went to college. That said — strong Python, real hands-on time with LLM APIs, and a working understanding of RAG, embeddings, vector stores, and agent frameworks (LangGraph, CrewAI, custom — we don't care which, we care that you've debugged one in production).

Bonus points if you've touched voice stacks (ElevenLabs, Deepgram, LiveKit, Retell, or built your own pipeline), evals frameworks, or call analytics tooling.

The non-negotiable: you've shipped something a real user used. A side project with 100 users teaches you more than a paper with 1,000 citations. We want people who know the difference between a demo that works in a notebook and a system that survives Tuesday morning traffic in Coimbatore.

What will make you struggle here

If you need clean specs, you'll be unhappy. If you think prompt engineering is beneath you, you'll be unhappy. If you want to stay in the model layer and never look at a CRM, you'll be unhappy. Our best engineers spend half their time in code and half their time on calls with sales managers in Tier-2 cities figuring out why a customer hung up.

The stack you'll touch

LLMs (frontier + open), RAG over our internal car/customer/policy corpus, voice agents on Indian telephony, real-time evals on AssemblyAI transcripts, integrations across our CRM, inventory, and scheduling systems.

Qualifications

B.Tech / B.E. preferred but not required. Show us what you've built.