Senior Staff Software Engineer-Voice AI Agent Prompting

Description

The company builds voice AI agents whose behavior comes from models rather than hand-written call-flow logic. These agents understand what a caller wants over live audio, reason over context, call tools, and take actions for the caller.

You take a role across the whole lifecycle: turning unclear call flows and telephony needs into precise specs, writing the instructions and prompts that steer the agent, and showing through automated evaluation that it behaves correctly on real calls.

This is not an ML or AI research role, and you will not train foundation or speech models. It also differs from conventional software work, since the system picks its execution path at runtime instead of following fixed call-flow logic.

Voice makes the job harder in two ways. Much of the logic lives in natural language, which needs the same versioning and review as code, and it must cope with turn-taking, barge-in, and speech recognition errors. Because behavior is probabilistic and happens live, correctness is measured at scale, covering latency, containment, and transcription accuracy. You are accountable for what reaches a live caller, whether a person or an agent wrote it.

Responsibilities

• Build agentic voice applications that interpret intent, reason across several steps, invoke tools, and act for the caller, along with the CTI, CCaaS, and telephony integrations needed to run them in production.
• Break call flows and business processes into steps a voice agent can carry out. Decide where it should act alone, where a live agent must take over, and how misrecognition, retries, and escalation are handled.
• Design for voice: keep context across turns, manage barge-in and interruptions, recover from speech recognition errors, clarify caller intent, and confirm before consequential actions, all within the latency a live call allows.
• Write, structure, and version the system instructions, role definitions, tool descriptions, guardrails, and escalation rules. Treat them as engineering artifacts with review and regression tests.
• Tune prompts against measured call results rather than how they look in a demo. Cover task decomposition, golden call transcripts, structured output schemas, grounding and citation, handing off to a human, and the token and latency cost that affects the caller.
• Build and run the evaluation that keeps changes safe on a live channel: golden call datasets, multi-turn suites, model-as-judge scoring checked against human QA, CI gates, and drift detection. Also test for adversarial inputs, jailbreaks, grounding, and tool selection specific to voice.
• Turn customer call-flow and telephony requirements into testable specs with clear scope, constraints, and acceptance criteria. Split the work into tasks sized for AI coding agents, then review their output for correctness and maintainability. You answer for the result, whoever or whatever produced it.
• Work as a forward deployed engineer inside a customer's telephony, CTI, and CCaaS setup. Use their call data, integrations, and channels, tune instructions and evaluation sets against live traffic, and bring back evidence that improves the product.
• Monitor containment, hallucination, unsafe actions, and transcription accuracy on live calls. Guard against prompt injection and data leakage, and feed production failures back into specs and evaluation sets.
• Work with product managers, designers, and platform teams to set success criteria, make tradeoffs, and explain capability and risk clearly.

Requirements

• At least 10 years of software engineering experience, including at least 2 years deploying contemporary AI solutions at customers.
• A record of building and running production voice applications that real callers depend on, not demos.
• Hands-on experience writing agent instructions and prompts for voice, designing AI-driven autonomous call workflows, and building evaluation that checks them against live traffic.
• Willingness to work directly with customers as a forward deployed engineer, embedded in their voice and contact center environment.
• A bachelor's degree in computer science, software engineering, or a related technical field, or equivalent practical experience.
• Solid grounding in data structures, algorithms, system design, APIs, data modeling, and testing.
• Experience building AI-native voice applications: agent orchestration, tool calling, context assembly, grounding on enterprise data, and handling the latency, cost, and failure modes of live audio.
• Experience designing autonomous multi-step voice workflows with decision points, autonomy limits, human checkpoints, retries, and observability into what the agent did and why.
• Practical work with telephony integration, speech recognition and synthesis, and streaming audio, including barge-in, turn-taking, and confirmation patterns.
• Skill in writing and maintaining instructions, role definitions, tool descriptions, guardrails, and refusal and escalation rules, with versioning, review, and regression coverage.
• Intent-driven prompt design for voice, using decomposition, golden and few-shot examples, schema enforcement, grounding, and iteration based on call-level results.
• Ability to build measurable frameworks for call quality, agent behavior, tool selection, and regression risk, including model-as-judge scoring calibrated by human QA, and adversarial, jailbreak, and grounding tests.
• Ability to define call-flow problems precisely enough for another engineer or an AI agent to implement them correctly, and to judge whether a problem belongs in code, in instructions, or with an agent.
• Working knowledge of prompt injection, call-recording and sensitive-data leakage, over-broad tool access, and unsafe autonomous action, and the ability to turn these into guardrails, least-privilege controls, and monitoring.
• Comfort working inside customer telephony environments with incomplete information and fixed deadlines, and the ability to explain capability, limits, and risk to non-engineers.
• Desired: prior work as a forward deployed, solution, or professional services engineer delivering software inside a customer's environment.
• Desired: telephony, contact center, IVR, speech recognition and synthesis, streaming audio, or real-time latency work.
• Desired: experience working with conversation or content designers on dialogue flow, tone, and error recovery.
• Desired: experience with evaluation frameworks, prompt and instruction management tools, LLM tracing, or analysis of production call transcripts at scale.
• Desired: domain depth in customer service, contact center operations, sales, or enterprise workflow, enough to challenge a requirement and not only build it.
• Desired: advanced degrees or certifications. These are a plus but do not replace a proven record of shipping reliable production voice AI agents.

Conditions

Base pay for this role in this location is $190,900 to $334,100, plus equity when applicable, variable or incentive compensation, and benefits.

Total pay depends on factors such as qualifications, skill level, competencies, and work location, and the base pay shown is a guideline. Pay is set by the geographic location of the role and may change with it.

Sales positions generally include an On Target Earnings incentive structure.

Work persona (flexible, remote, or required in office) is assigned based on the nature of the work and the assigned work location. The company may confirm the distance between your home and the nearest office with a third-party service.

Benefits

• Health plans, including flexible spending accounts.
• 401(k) plan with company match.
• Employee stock purchase plan (ESPP).
• Matching donations.
• Flexible time away plan.
• Family leave programs.