SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 190,900 - 334,100 / annual
ServiceNow is seeking a Senior Staff Software Engineer to design, build, ship, and operate voice AI agents whose behavior is model-driven rather than explicitly authored. This role bridges conversational AI and production telephony, requiring deep expertise in agentic systems, prompt engineering, and voice-specific challenges.
You will work across the full lifecycle of voice AI agent development: translating ambiguous call flows and telephony requirements into precise specifications; authoring the instructions, prompts, and guardrails that govern agent behavior; and proving correctness through automated evaluation against real call traffic. This is distinct from traditional software engineering (deterministic call flows) and from AI research (you do not train foundation or speech models). Instead, you engineer systems where behavior is probabilistic and happens live on calls, making correctness a measurement problem rather than an assertion problem.
Key responsibilities include:
- Design and ship agentic voice applications with intent interpretation, multi-step reasoning, tool invocation, and action on the caller's behalf, integrated with CTI, CCaaS, and telephony platforms.
- Decompose call flows and business processes into steps and decision points a voice agent can execute, determining autonomy boundaries and human-agent handoff points.
- Build experiences that hold context across turns, manage barge-in and interruption, recover from speech-recognition errors, and disambiguate intent within live-call latency budgets.
- Author and maintain voice agent instructions as engineering artifacts: system prompts, role definitions, tool descriptions, guardrails, and escalation rules under version control and regression coverage.
- Engineer prompts for reliability: iterate against measured call outcomes using task decomposition, golden transcripts, structured schemas, grounding, and token/latency optimization.
- Design and operate automated evaluation frameworks for non-deterministic voice behavior: golden call datasets, multi-turn suites, model-as-judge scoring calibrated to human QA, CI gates, drift detection, and adversarial testing.
- Convert customer requirements into testable specifications with explicit scope and constraints; decompose work into agent-sized tasks; own correctness regardless of what produced the output.
- Work as a forward deployed engineer embedded in customer telephony and contact center environments, tuning instructions and evaluation against live traffic.
- Monitor call containment, hallucination, unsafe actions, and transcription accuracy; defend against prompt injection and data leakage; feed production failures back into specifications and evaluation.
- Collaborate with product, design, and platform teams to define success criteria and communicate capability and risk.
Requirements:
- Minimum 10 years of software engineering experience with at least 2 years deploying contemporary AI solutions at customers.
- Demonstrated track record building, shipping, and operating production voice applications with hands-on delivery of AI-native voice agent features that real callers depend on (not demos).
- Direct experience authoring agentic instructions and prompts for voice, designing AI-driven autonomous call workflows, and building evaluation/testing against live call traffic.
- Willingness to work directly with customers in a forward deployed capacity, embedded in their voice and contact center environment.
- Bachelor's degree in Computer Science, Software Engineering, or related technical field, or equivalent practical experience.
- Strong software engineering fundamentals: data structures, algorithms, system design, APIs, data modeling, testing.
- Hands-on experience building AI-native voice applications: agent orchestration, tool/function calling, context assembly, grounding against enterprise data, handling latency/cost/failure modes of live audio.
- Demonstrated experience designing workflows where voice agents execute multi-step call processes: call-flow decomposition, decision points, autonomy boundaries, human-in-the-loop checkpoints, exception handling, observability.
- Practical experience with telephony integration, speech recognition/synthesis, streaming audio, barge-in, turn-taking, confirmation patterns.
- Demonstrated skill writing and maintaining natural-language logic governing voice agent behavior: instructions, role definitions, tool descriptions, guardrails, refusal/escalation rules, with versioning and review discipline.
- Intent-driven prompt design tuned for voice: decomposition, golden/few-shot examples, structured output/schema enforcement, grounding/citation, disciplined iteration against call-level evaluation.
- Ability to build measurable frameworks assessing call quality, agent behavior, tool-selection accuracy, regression risk: golden call datasets, scenario suites, model-as-judge scoring with human calibration, continuous evaluation, drift detection, adversarial/jailbreak/grounding testing.
- Ability to define call-flow problems rigorously and decide when to solve in code, instructions, or via agent delegation.
- Working knowledge of AI-integrated voice security risks: prompt injection, sensitive-data/call-recording leakage, over-broad tool access, unsafe autonomous action, translated into concrete guardrails and monitoring.
Desired experience:
- Prior work as a forward deployed engineer, solution engineer, or in professional services delivering software inside customer environments.
- Telephony and contact center platforms, IVR, speech recognition/synthesis, streaming audio, real-time latency optimization.
- Experience working alongside conversation or content designers on dialogue flow, tone, error-recovery design.
- Evaluation frameworks, prompt/instruction management tooling, LLM application tracing, production call transcript analysis at scale.
- Customer service, contact center operations, sales, or enterprise workflow domain depth sufficient to challenge requirements.
