Your opportunity
Our client is a small, talent-dense AI product team building conversational, agentic systems for emergency response. The team began as a seed-stage startup focused specifically on public-safety communications and was acquired in 2026 by a global public-safety technology company, where it now operates as a specialized division within its broader AI organization.
Their initial wedge is deceptively simple: public-safety answering points are chronically understaffed, yet highly trained 911 telecommunicators spend meaningful amounts of time handling calls that are not emergencies. The platform uses voice AI to autonomously handle designated non-emergency interactions, gather information, answer questions, route callers and escalate appropriately, allowing human operators to concentrate on situations where their judgment matters most.
The startup proved that use case independently. The acquisition gives the team something much harder for an early-stage company to manufacture: distribution into established emergency-response infrastructure and the opportunity to integrate its technology across a much broader suite of command-centre products. The group is now a strategically important part of a larger effort to introduce useful, human-centred AI throughout emergency response (from the initial call through dispatch and field operations).
The engineering environment has retained much of its startup character. The team remains relatively small, engineers have substantial ownership, and the company is growing the group materially following the acquisition. This role sits in a division focused on the systems that make the voice agent itself work rather than primarily building downstream customer integrations.
The team actively prototypes with emerging technologies (voice-to-voice models) and new inference architectures, but a compelling demo is not sufficient reason to rebuild a mission-critical system. Public-safety customers need reliability, auditability and the ability to understand why a system behaved the way it did. Today, the platform deliberately separates speech-to-text, reasoning and text-to-speech into modular components, allowing individual providers to be changed or failed over independently rather than making the entire product dependent on a single inference provider.
That means some of the most interesting engineering work lives in the space between the models: orchestrating real-time conversations, managing latency and interruption, integrating deterministic software with probabilistic systems, designing graceful failover, tracing model behaviour and maintaining a product that still works when an upstream AI provider does not.
As a Senior Software Engineer, you’ll help set technical direction for this platform and own difficult problems across real-time AI pipelines, backend systems, telephony and production infrastructure. You’ll also join a team that encourages engineers to use AI agents in their development workflows wherein humans remain accountable for quality, judgment and the software operating in production.
Key responsibilities
• Voice AI platform engineering: Design, build and operate the real-time systems that power conversational AI experiences
• Software architecture: Set technical direction across complex backend and distributed systems, making considered trade-offs across latency, reliability, maintainability, security and performance
• Real-time systems: Solve difficult problems involving streaming audio, concurrency, interruption handling, asynchronous workflows and low-latency communication
• Agentic AI: Build and evolve production systems that combine LLMs, deterministic software, tool calling and human-in-the-loop/HITL workflows
• Reliability & resilience: Design systems that tolerate failures across external inference providers, infrastructure and network dependencies without turning every upstream outage into a customer outage
• Evals & observability: Improve the testing, evaluation, tracing and monitoring systems used to understand both conventional software behaviour and the less deterministic behaviour of AI components
• Infrastructure & operations: Contribute across cloud architecture, deployment, observability, incident response, performance optimization and the operational practices required for mission-critical software
• AI-augmented engineering: Deploy contemporary coding agents and AI development tools to increase development velocity while maintaining a high bar for testing, review, maintainability and human judgment
• Technical leadership & mentoring: Guide architecture and design decisions, conduct thoughtful reviews and help less experienced engineers develop stronger technical judgment
• Cross-functional problem solving: Work closely with product, operations and engineering leadership to translate messy customer and operational requirements into durable technical systems
• Technical credibility: Participate in conversations with customers, partners and stakeholders when a problem requires someone who understands the underlying systems deeply
Technical ecosystem
The platform combines conventional distributed software, real-time communications infrastructure and a modular voice-AI pipeline. The technologies below represent the broader system engineers are building within rather than a checklist of tools every engineer is expected to use day to day.
• Voice & AI inference: Deepgram (speech-to-text), OpenAI (LLMs), ElevenLabs (text-to-speech)
• Real-time AI orchestration: LiveKit
• Application & systems engineering: Elixir, Go, Python, Rust
• Real-time communications: WebRTC, WebSockets, RTP/SRTP, media streaming and low-latency audio transport
• Telephony & voice infrastructure: PSTN, SIP trunking, PBX systems, call routing
• Database: PostgreSQL
• Cloud & deployment: Azure, with AWS/GCP experience also relevant; containerized application environments
• Observability & AI tracing: Datadog, Langfuse
• Incident management: incident.io
The stack is intentionally modular. Individual speech, inference and voice providers can be substituted or failed over independently, and the team expects its technology choices to continue evolving as the underlying capabilities improve. They’re well aware of the frontier capabilities (such as GPT-Live) and are constantly evaluating them against their use case.
Your know-how
• You have 7+ years of production software engineering experience and a track record of taking ownership beyond the boundaries of individual tickets or features
• You have designed, shipped and operated complex backend, distributed or real-time systems where reliability and performance genuinely matter
• You have experience integrating LLMs, agentic systems or other AI capabilities into production software (and understand the gap between making an AI feature work in a prototype and making it trustworthy in production)
• You can reason comfortably about concurrency, asynchronous systems, latency, failure modes and the behaviour of systems composed of multiple external dependencies
• You have strong software engineering fundamentals and care about tests, observability, clean abstractions, security and maintainability even when development velocity is high
• You use AI development tools in your own work and understand how to delegate aggressively to agents without delegating engineering judgment
• You can move between architecture and implementation detail, contributing directly to difficult code while also helping the broader team make better technical decisions
• You are comfortable working with ambiguous product and operational problems where the technical requirements need to be discovered rather than simply implemented
• You communicate effectively with other engineers, product and operational stakeholders and have a talent for explaining the behaviour of complicated systems
• You mentor effectively and raise the quality of the engineers and systems around you
• You are motivated by building AI systems whose value extends meaningfully beyond a demo or novelty use case
It’s a bonus if
• You have built or operated production voice-AI systems (particularly using LiveKit, Pipecat or an equivalent real-time communications framework)
• You have experience with speech-to-text/STT, text-to-speech/TTS, voice-activity detection, endpointing, interruption handling, diarization, noise filtering or real-time audio streaming
• You have experience with telephony infrastructure including PSTN, SIP trunking, PBX systems or call routing
• You have worked with WebRTC, WebSockets, RTP/SRTP or other real-time media and communication protocols under production load
• You have designed highly concurrent or fault-tolerant services using Elixir, Erlang, Go, Rust or another language well suited to those problems
• You have implemented evaluation frameworks, prompt test harnesses or production observability for LLM or agentic systems
• You have designed systems that gracefully fail over between multiple infrastructure or AI providers
• You have worked in public safety, healthcare, defence, financial infrastructure or another domain where reliability, security, auditability and human judgment carry unusually high stakes
Interested in learning more?
Please upload your resume or a PDF export of your LinkedIn profile to talent@lutrapartners.com with “Senior Software Engineer, Voice AI” as the subject line. One of our partners will be in contact shortly.
C ompensation
The base pay range for this role in Canada is CA$130,000 – CA$220,000 per year.
The base pay range for this role in the United States is $110,000 – $200,000 per year.