Senior Staff Software Engineer_Voice AI Agent Prompting

The Forward Deployed Engineer, Voice AI Agent Engineering designs, builds, ships, and operates voice AI agents whose core behavior is model-driven rather than explicitly authored — conversational and agentic experiences that interpret caller intent over a live audio channel, reason over context, invoke tools, and act on the caller's behalf. The engineer works across the full lifecycle: shaping ambiguous call flows and telephony requirements into precise specifications, authoring the instructions and prompts that govern voice agent behavior, proving through automated evaluation that the result behaves correctly on real calls. This is not a machine learning or AI research role — the engineer does not train foundation or speech models — but it is distinct from traditional software engineering, where systems follow deterministic call-flow logic rather than selecting execution paths at runtime.  Two things follow, and voice makes both harder. First, the agent's most important logic often lives in natural language — instructions, prompts, tool descriptions, and guardrails — which must be engineered, versioned, and reviewed with the same discipline as code, while also holding up under what a spoken channel adds: turn-taking, barge-in, and disambiguating what speech recognition got wrong. Second, because behavior is probabilistic and happens live on a call, correctness is established by measuring behavior at scale — including latency, containment, and transcription accuracy — rather than by asserting fixed outputs, making automated evaluation a first-class engineering activity rather than a quality-assurance afterthought. The engineer owns the correctness of what ships to a live caller, whether a human or an agent produced it.  What You Get to Do in this role:  • Build voice AI agents. Design and ship agentic voice applications — intent interpretation, multi-step reasoning, tool invocation, and action on the caller's behalf — together with the CTI, CCaaS, and telephony integrations that make them usable in production.  • Design AI-driven autonomous voice workflows. Decompose call flows and business processes into the steps and decision points a voice agent can execute — determining where autonomy is appropriate, where a live-agent handoff is required, and how misrecognition, retries, and escalation are handled.  • Engineer for what voice demands. Build experiences that hold context across turns, manage barge-in and interruption, recover gracefully from speech-recognition error, disambiguate caller intent, and confirm before consequential actions — all within the latency budget a live call imposes.  • Author and maintain voice agent instructions. Write, structure, and version the system instructions, role definitions, tool descriptions, guardrails, and escalation rules that govern voice agent behavior — as engineering artifacts under review and regression coverage, not configuration text.  • Engineer prompts for reliability rather than demo quality. Iterate prompts against measured call outcomes: task decomposition, golden call transcripts, structured output schemas, grounding and citation, graceful hand-back to a human agent, and the token and latency cost that directly shapes caller experience.  • Build automated evaluation and test non-deterministic voice behavior. Design and operate the evaluation that makes change safe on a live channel — golden call datasets, multi-turn conversation suites, model-as-judge scoring calibrated to human QA review, CI gates, and drift detection — plus adversarial, jailbreak, grounding, and tool-selection testing specific to voice.  • Specify precisely and direct AI coding agents. Convert customer call-flow and telephony requirements into testable specifications with explicit scope, constraints, and