Position : Full-Stack Developer
Employment Type: Full-time
Location: On-site (Ahmedabad)
Job Description
Full-Stack Developer
The Work:
We run live voice AI conversations in production every day. A delay over 600ms makes the AI feel broken. A dropped webhook means a CRM record never gets updated and a real customer follow-up never happens. A misrouted call costs our client a deal.
That’s the kind of system you’d be building.
Concretely, you’ll work on:
- Real-time voice pipelines - WebSocket and streaming audio, turn-taking, barge-in, latency budgets that actually matter
- Conversation orchestration - stateful call flows, tool-calling, interruption handling, recovery from partial failures
- CRM and telephony integrations - HubSpot, Salesforce, Zoho, GoHighLevel, Twilio, Plivo. OAuth flows that break, webhooks that retry, signatures that need verifying
- High-write data flows - call events, transcripts, audit trails - across MongoDB, PostgreSQL, and Redis
- Queue-based and event-driven workflows - long-running call → transcription → CRM update → follow-up chains
- Deploys, monitoring, and incident response on Docker and Kubernetes
The Stack:
Non-negotiable in production:
- Node.js + TypeScript
- WebSockets / real-time bi-directional systems
- Redis
- One of MongoDB or PostgreSQL (you'll learn the other here)
- Docker
- Git, CI/CD
Must have used, depth flexible:
- Next.js (React, TypeScript)
- Queues - BullMQ, RabbitMQ, Kafka, or similar
- Kubernetes
- Microservices or service-oriented architecture (built or worked inside, not just monoliths)
You’ll be exceptional here if you’ve also done:
- Real-time audio or voice AI systems
- Telephony - Twilio, Plivo, Asterisk, FreeSWITCH, WebRTC, SIP
- STT / TTS pipelines - Deepgram, ElevenLabs, Whisper, Azure Speech
- Webhook reliability patterns - idempotency, dead-letter queues, replay protection
- Observability - structured logging, metrics, distributed tracing
- LLM and prompt engineering work in production
We value depth over years. Three years deep in production microservices is more interesting to us than seven years of internal-tool work.
How We Work:
- You'll work directly with engineering leadership and the founding team
- We ship fast, but we expect tests, code reviews, and post-incident write-ups
- We flag slipping work before the deadline, not after
- We care about per-call infrastructure cost - engineers here think about token spend, STT/TTS minutes, and inference cost as part of system design
- We're early. Tooling is still being built. You'll sometimes write infrastructure a larger company would have a platform team for. If you want clean abstractions handed to you, this isn't the role.
Note: On-site means you’re in the room with the team. Decisions happen fast. So does shipping.
