Senior Platform Engineer, Voice AI

About the Role Together AI is building the best inference infrastructure for voice applications. Our Voice AI platform powers production-grade, real-time voice agents and applications — serving speech-to-text and text-to-speech models with best-in-class latency and reliability. We're looking for a Senior Platform Engineer to own the API and infrastructure layer for voice workloads. You'll build the real-time WebSocket and HTTP APIs that developers use to ship voice experiences, design autoscaling for latency-sensitive streaming workloads, and ensure our multi-provider voice platform is reliable enough for production voice agents handling millions of calls. This is a foundational hire on a small, high-impact team. Voice APIs have fundamentally different infrastructure requirements than text-based inference — bidirectional audio streaming, stateful connections, tight latency SLOs, and complex multi-model routing. You'll define how developers interact with Together's voice platform as we grow from early customers to the default infrastructure for voice AI. Own the real-time API layer (WebSocket + HTTP streaming) that powers Together's voice platform. Design autoscaling and orchestration for voice workloads running on tens of thousands of GPUs. Build the developer experience — APIs, observability, and tooling — for a fast-growing product area. Work with production voice customers (contact centers, AI agents, communication platforms) to ship what they actually need. Join a small, early-stage team with outsized impact on a new product line. Responsibilities Build and harden real-time WebSocket and HTTP streaming APIs for STT and TTS — including connection lifecycle management, backpressure, error handling, and reconnection, at the reliability bar needed for production voice agents. Design and ship autoscaling for voice model endpoints that handles bursty, real-time traffic patterns — accounting for concurrent connection limits, streaming state, and hard latency ceilings. Implement voice-specific API features: word-level alignment, speaker diarization in realtime, audio format flexibility (g711/mulaw for telephony, PCM, WebRTC formats), pronunciation controls, and multi-context WebSocket support. Build voice-specific observability — latency breakdowns, audio quality signals, and dashboards that help both the team and customers debug issues. Own multi-model normalization across our model partners (Cartesia, Deepgram, Rime, and others), ensuring consistent API behavior regardless of the underlying provider. Collaborate with the ML engineering side of the team on the interface between the API layer and the model serving stack, ensuring latency and reliability requirements are met end-to-end. Contribute to developer experience — API design, documentation, integration cookbooks, playground and showcasing how best-in-class voice agents are built. Lay the groundwork for multiple new products down the line. Requirements 5+ years of experience building large-scale, real-time distributed systems and API services. Deep expertise in real-time streaming infrastructure — WebSocket server architecture, Server-Sent Events, bidirectional streaming, connection multiplexing, and stateful protocol design. Expert-level programming in TypeScript and Python; experience with Rust is a plus. Strong distributed systems fundamentals: load balancing, autoscaling, rate limiting, and traffic shaping for latency-sensitive workloads. Experience with Kubernetes — including custom autoscalers, resource management, and health checking for stateful services. Strong product sense — you care about API ergonomics and think about what developers building voice apps actually need. Comfort working on a small, early-stage team where you'll wear multiple hats and move fast. Experience with audio or media protocols (WebRTC, g711, PCM encoding) is a strong plus. Familiarity with ML model serving infrastructure and how inference engines work is a plus — you'll interface with the