Senior Voice AI Engineer

Full-Time
Fully_Remote
Argentina, Chile, Colombia, Mexico, Peru, Uruguay, Brazil

Summary of the Role

We are looking for a Senior Voice AI Engineer to own a zero-to-one, multi-tenant inbound voice reception and case intake platform for law firms, from architecture through production deployment. In this hands-on applied engineering role you will build the real-time audio pipeline for inbound calls over SIP/PSTN, tune voice mechanics (barge-in, turn-taking, VAD, latency), design secure tenant isolation and integrate deterministic workflows with legal CRMs and calendars, with the clear goal of shipping a live, pilot-ready MVP within 90 days.

Desired Skills

  • Full-Time availability
  • Advanced English (C1) or above
  • 5+ years of professional software engineering, with deep experience in system architecture, deployment and testing
  • Production Voice AI: You have built and launched at least one phone-based, real-time voice AI agent in production or in a high-stakes customer pilot (demo and hackathon projects do not count).
  • Advanced Asynchronous Python: Expert asyncio, streaming architectures, WebSockets, API development (FastAPI, Pydantic) and distributed debugging.
  • Voice AI Frameworks: Hands-on experience with LiveKit Agents or comparable frameworks (Pipecat, Vapi, Retell, or custom WebRTC/WebSocket audio loops).
  • Telephony & Audio Engineering: Solid practical knowledge of SIP, PSTN providers, codecs, latency tuning, VAD, turn-taking and barge-in.
  • Deterministic Agent Architecture: Proven ability to build structured outputs, explicit state machines, schema validation and guardrails that prevent hallucination and unauthorized legal advice.
  • Multi-Tenant System Design: Experience designing secure multi-tenant SaaS platforms with strict data segregation, configuration versioning and auditable logging.

Responsibilities

  • Real-Time Voice Pipeline: Design, build and optimize the end-to-end streaming audio pipeline (LiveKit Agents, Pipecat or WebSockets), integrating streaming STT (Deepgram), LLMs (OpenAI Realtime, Anthropic, Bedrock) and low-latency TTS (ElevenLabs, Cartesia) or Speech-to-Speech models.
  • Conversational Mechanics & Audio Tuning: Implement robust VAD, silence handling, interruption recovery, natural turn-taking, endpointing and barge-in that hold up on noisy lines, varied accents and mid-sentence corrections.
  • Deterministic Workflows & State Machines: Engineer state management, retry logic, timeouts and idempotent tool calling so critical intake steps (consultation booking, warm transfers, structured field extraction) do not rely on raw prompt generation alone.
  • SIP & Telephony Infrastructure: Configure and manage inbound SIP trunks, phone numbers, DTMF detection, call routing and warm transfers with carriers (LiveKit, Twilio, Telnyx, Bandwidth), monitoring call drop rates and carrier failovers.
  • Multi-Tenant Configuration & Security: Build a reusable, versioned configuration model so practice areas, qualification criteria, disclosures and transfer rules scale to new firms through configuration instead of code, while enforcing tenant isolation, encryption, access controls and audit logs.
  • Legal Integrations: Integrate deterministic intake workflows with legal CRMs and calendars (Clio, Filevine, Lawmatics).
  • Automated Voice Evals & Testing: Build automated conversation test suites that simulate noisy backgrounds, interruptions and edge-case caller behavior, tracking intake completeness, field accuracy, latency and cost per call.

Nice-to-Haves

  • LiveKit & Agent Frameworks: Direct experience with LiveKit (Agents, Cloud, SIP) or Strands Agents.
  • Realtime Models: Exposure to AWS Bedrock (Nova Sonic) or the OpenAI Realtime API.
  • Domain Experience: LegalTech (Clio, Filevine, Lawmatics) or other privacy-regulated environments such as healthcare or financial services.
  • Multilingual & Omnichannel: Multilingual voice agents or omnichannel voice-to-chat design.
  • Platform & Observability: PostgreSQL, PII redaction, Docker, Kubernetes/Cloud Run, CI/CD, Grafana, OpenTelemetry and Prometheus; STT providers such as AssemblyAI.

Who You Are

  • High-Agency Zero-to-One Builder: You take ambiguous business goals and turn them into working software under rapid feedback cycles.
  • Pragmatic Applied Engineer: You focus on shipping reliable, low-latency value to customers rather than chasing academic AI novelties.
  • Architecturally Rigorous: You obsess over deterministic failure recovery, data isolation and auditable system behavior.

Some benefits:

  • 🏢 Offices in some cities
  • 🖥️ 100% remote work
  • ⌚ Full-time schedule, flexible according to objectives
  • 🏖️ PTO & holidays
  • ⚕️ Medical insurance

About Howdy

Howdy.com, founded in 2018 and headquartered in Austin, Texas, helps US companies who want to hire, manage, and retain their teams in Latin America (LatAm) directly but need help with multinational logistics, contracts, compliance, and culture. Companies that use Howdy.com get the best talent available in LatAm and gain access to an entire network and a thriving community of professionals who are changing the world. By partnering with Howdy.com, companies can expand their physical presence into some of the fastest-growing economies in LatAm.

Howdy.com is a member of Y Combinator and has garnered significant support from prominent investors, including Greycroft and Obvious Ventures. The company raised over $20 million in a series A venture capital round.

Our core values

#1 Sports Team: At Howdy, we win together. From players to support, everyone is vital to our success. We hire for excellence, prioritize teamwork, and strive for continuous improvement. We collaborate, seek advice, and actively contribute to Howdy's victories.

Altruism: Demonstrating altruism involves prioritizing the team and assuming the best in others. We communicate openly, provide honest feedback, and extend grace. Altruism is selfless service, focusing on supporting our players and team growth.

Curiosity: Being curious at Howdy means having the willingness to learn, adapt, and explore new ideas. We question existing beliefs, embrace humility, and see curiosity as our superpower. Demonstrating curiosity involves researching unfamiliar tasks, asking questions to understand the full picture, and seeking better ways to complete routine tasks.

Have Spirit: Having spirit at Howdy is about celebrating wins, building a sense of community, and bringing positivity. Demonstrating spirit involves attending events, getting to know teammates, participating in challenges, and proudly wearing the Howdy swag. Simply put, it's about bringing a super-fan spirit to work every day.