- Salary
- ₹30 – ₹55
- Location
- Gurgaon, India
- Department
- IT
- Seniority
- Senior
- Source
- Keka
Description
About Spyne
Spyne is building the AI operating system for automotive retail. Starting with visual merchandising and expanding into conversational AI, Spyne helps dealers move inventory faster, respond to buyers instantly, and run more of the dealership stack with AI.
Founded by operators from Amazon, Cox, and OYO, Spyne has grown 30x in under three years and now powers 1,500+ dealerships across the US. Its products, Studio AI and Vini AI, are coming together into a full-stack Retail AI platform for modern dealerships.
Backed by Accel, Vertex Ventures, and Storm Ventures, Spyne is emerging as one of the category leaders redefining how dealerships operate in the AI era.
- Studio AI: https://www.youtube.com/watch?v=j0_CsQpCbwg
- Vini AI: https://www.youtube.com/watch?v=WFjhPWp6gJU
The role
You will lead the core of Conversational & Voice AI platform, Vini which handles inbound and outbound calls for dealer BDCs across multiple dealer workflows.
- Sales: lead follow-up, qualification, and appointment setting.
- Service: service appointment booking, recall outreach, status updates.
- Acquisition: outbound to source trade-in inventory from past customers.
- Finance: financing, lease, emi, credits across inventory, customer, and dealer.
- The agent does the work a human BDC rep does. On the phone. In real time. With DMS, IMS, and CRM context. Our product goal is to make the human BDC seat redundant.
What you’ll own
- The Vini product roadmap across Sales, Service, and Acquisition. You decide what we build, in what order, and why.
- Agent design. The prompt stack, tool and function design, conversation flow, escalation logic, guardrails, voice persona. You own how the agent behaves and how we make it better.
- Evals. Multi-turn conversation quality, function calling accuracy, interruption handling, turn taking, ASR error recovery, hallucination rate, task completion. You define the metrics, build the test sets, and run regression on every release.
- The latency budget. TTFT, endpointing, tool call latency, end-to-end response time. You own perceived response time.
- Model and vendor choices. ASR (Deepgram, AssemblyAI, in-house), TTS (ElevenLabs, Cartesia, others), LLM provider mix per use case. You make the calls based on cost, latency, and quality data.
- Customer deployment quality. You sit with deployment engineers and Customer Success on the hardest 20 accounts. You listen to call recordings. You find the failure modes before customers do.
- The dealer-facing configuration surface. The UI that lets a non-technical dealer onboard, edit prompts, change business hours, and read transcripts.
- DMS and IMS integrations roadmap that unlocks the agent's capability ceiling: CDK, Reynolds & Reynolds, Tekion, vAuto, and the rest.
What we expect
- 5+ years in product management. At least 2 of those years on a voice or conversational AI product that real customers used at real volume. Demos and internal tools do not count.
- You have personally owned a production metric on a conversational agent: containment, resolution rate, AHT, CSAT, transfer rate, or task completion. You can explain in 5 minutes how you moved it and what you tried that did not work.
- You have built evals for non-deterministic systems. Golden sets, simulation harnesses, LLM-as-judge with human-graded calibration, regression suites that run on every prompt change.
- You have made architecture calls on STT, LLM, and TTS pipelines. You know what TTFT, endpointing latency, and barge-in mean and you have tuned them. You know why function calls stack latency and what to do about it.
- You have shipped agentic systems with tools and integrations. You know why instruction following breaks down in multi-turn and what your options are.
- You write your own SQL or use a notebook. You do not file tickets to the data team to check call outcomes.
- You have run customer deployments yourself. You have been the PM on a call with an angry customer at 9pm on a Friday. You know what edge cases look like in production.
- You write clearly. Short PRDs. Specific success criteria. No fluff.
Bonus:
- Contact center, BDC, or call ops background.
- Auto retail context: dealerships, DMS, F&I, fixed ops, used car operations. If you have not worked in auto, you will learn it inside 60 days.
- CS or engineering degree, or you have shipped code in the last 2 years.
- 0 to 1 experience. You founded something, or you were the first PM at a startup.
- Bilingual or multilingual product experience (Spanish, French Canadian).
Skills
SQLMerchandisingVisual Merchandising