Backend Engineer, AI Voice Platform (Python, Kubernetes)

NeuroDrift is a US-based voice AI company. We build AI voice agents that answer real phone calls for enterprise customers. The platform is Python, and it sits between carrier telephony and speech and language models. It runs live conversations under tight latency budgets and hands calls to human agents when needed. It is in production today and growing.

You will own a large part of that platform: the Python services behind it, how they run on Kubernetes, and how they integrate with speech and LLM providers. You will ship features, evaluate new models as they arrive, and be the person who understands why the system behaves the way it does on a real call.

You do not need a voice or WebRTC background. We are looking for a strong Python backend engineer with real infrastructure skills and the curiosity to go deep into a new domain. The telephony and real-time audio parts are learnable.

What you'll do

•⁠ ⁠Design and ship features across a Python asyncio service, a Django backend and an LLM-driven conversation engine

•⁠ ⁠Run the platform on GKE: deploys, rollouts, observability, capacity

•⁠ ⁠Integrate and evaluate speech and LLM providers as the model landscape changes

•⁠ ⁠Debug across service boundaries: from a carrier trace to a Kubernetes scheduling decision to a line of Python

What we need

•⁠ ⁠3 to 5 years of production Python, including real asyncio experience (concurrent tasks, cancellation, streams) and the habit of reading a library's source when the docs run out

•⁠ ⁠Kubernetes at a working level: you have deployed and debugged services with kubectl and Helm and handled a rollout that went wrong

•⁠ ⁠One major cloud, GCP preferred: networking, load balancers, IAM

•⁠ ⁠Django or a comparable framework, PostgreSQL, Redis

•⁠ ⁠CI/CD with GitHub Actions or similar

•⁠ ⁠A diagnostic mindset: you test hypotheses against logs and data, and you say not sure yet rather than guess

Nice to have

•⁠ ⁠Telephony or real-time audio exposure: SIP, RTP, WebRTC, call flows

•⁠ ⁠LiveKit, Pipecat or a similar real-time voice framework

•⁠ ⁠Speech or LLM APIs in production

•⁠ ⁠Prompt and flow work on LLM-driven conversation systems

•⁠ ⁠Prometheus, Grafana, Terraform or OpenTofu

The setup

•⁠ ⁠Full-time, fully remote (India)

•⁠ ⁠Working hours 5:30 PM to 2:30 AM IST

•⁠ ⁠Immediate joiners

•⁠ ⁠Small team, direct communication, high ownership