Full-Stack Engineer, AI Systems
We are eight engineers building an AI sales coach that car dealerships put in the pocket of every person on their floor. This role owns the retrieval, agent and evaluation systems underneath it.
- Compensation
- $80,000 – $100,000
- plus equity
- Location and type
- Remote (US time zones) · Full time · Reports to the founders
What you will build
A retrieval layer per customer
Every dealership gets a model built from their own scripts, their sales process and recordings of their best closers, turned around in fourteen days. You own how that gets ingested, indexed and kept current as their process changes — and how we do it for the fiftieth customer as cheaply as the first.
Voice agents that hold a real sales conversation
A rep practises against an AI buyer with objectives, a difficulty setting and a specific objection to get past. Latency and realism are the product. If it feels like a chatbot, nobody opens it twice.
The grading harness
Every conversation is scored out of 100 across rapport, discovery, pitching value, objection handling and closing. Consistency matters far more than cleverness here: a score that moves when the input did not is worse than no score, and a sales manager will notice within a week.
Evals that let us ship on a Friday
We build something custom per customer, which means a prompt or retrieval change can quietly regress a dealership nobody was thinking about. You build the harness that proves it did not. This is the least glamorous item here and the one that decides how fast the rest moves.
Ingestion and analysis of real floor conversations
Reps record actual conversations with actual buyers. Transcripts come back, key moments get pulled out, weaknesses get turned into assigned practice automatically. Messy audio, accents, a showroom in the background.
The stack
- Language
- TypeScript end to end, strict.
- Backend
- Node services.
- Web
- Next.js App Router, React 19, Tailwind. This site and the manager dashboard.
- Mobile
- Two apps shipping in the App Store. The rep experience is mobile first.
- Infra
- Vercel.
- AI
- LLM providers, retrieval, speech. Specific choices are open and you will have a say in them.
What we need from you
- You are strong in Node and TypeScript and you have carried something to production in both.
- You have shipped retrieval-augmented generation that real users depend on, not a weekend demo. You have dealt with chunking, retrieval quality, and the afternoon it started returning confident nonsense.
- You have built agent harnesses: tool loops, retries, structured output, and the eval suite that keeps them honest.
- You are comfortable owning a system end to end on a team of eight engineers, with nobody to hand the ambiguous half to.
Helps, but we will not screen on it
- Realtime voice, or speech-to-text at volume.
- React Native.
- You have worked somewhere that sold to an industry you knew nothing about, and learned it.
Who you would be joining
CloserCoach started at the end of 2024 and went live in the middle of 2025. We have trained 15,000+ closers and recorded 1.1M minutes of real sales conversation. We raised just under a million dollars and spent it on engineers: twelve people, eight of them building.
Our customers are car dealerships, which is a better place to build AI than it sounds. Sales floors are measured constantly, the people on them are paid on results, and every conversation that matters happens out loud. You will talk to them. The last six demo calls are in our repo and engineers read them.
How hiring works
- Email us. A few lines on something you built with retrieval or agents, and what went wrong with it. A repo beats a CV.
- A call with a founder. Thirty minutes, no slides.
- A paid work session. A real problem from our backlog, a few hours, and we pay for your time.
- Offer.
hello@closercoach.ai · we reply to everyone.