AI Engineer, Realtime Voice Agent Platform

About the Role

Aloware is building a production-grade, agentic, S2S-first voice AI platform that answers and makes real phone calls at scale. Think of a platform competitive with Retell and Vapi, but built with an operator-grade mindset: low latency, configurable agent behavior, strong tool execution, and a reliable admin/control plane.

We’re forming a small, elite team of AI-enabled builders who use modern tools (e.g., Cursor, Claude Code) to deliver clean, maintainable, well-tested production code (not prototypes or AI-generated “slop”). You will work across the agent runtime and the admin platform, ship quickly, and set the quality bar for the entire system.

What You’ll Work On (Core Stack)

A real-time voice agent platform with a multi-tenant admin/control plane and a Python agent runtime:

• Admin / Control Plane: Next.js 16 + TypeScript dashboard backed by Supabase (Postgres + Auth + Realtime + RLS) for secure org-based multi-tenancy.
• Realtime Voice + Telephony: Self-hosted LiveKit for realtime voice + SIP (rooms, dispatch rules, SIP trunks) with Twilio as the PSTN bridge (numbers + TwiML webhooks routing calls into the realtime stack).
• Config + Workflows: Redis for low-latency agent configuration caching, synced from Postgres via durable background workflows (config sync, post-call analysis, webhook delivery, telephony resource sync).
• Agent Runtime: Python 3.13 agents using the LiveKit Agents SDK with plugins for turn detection/VAD, STT/TTS, LLM providers, and MCP for dynamic tool discovery and registration.
• Model Support: Both traditional STT→LLM→TTS and real-time / speech-to-speech (S2S) unified voice models for lower latency and more natural conversations
  .
  

Responsibilities

• Build a production-grade realtime voice agent platform: ship core call capabilities (e.g., transfers, voicemail/AMD-style handling), natural turn-taking, and reliable call flows that work at scale in real customer environments.
• Design and ship agentic behaviors: reliable tool execution, safe variable resolution, and orchestration of multi-step actions during live calls.
• Own the agent runtime and “in-call intelligence”: improve speech-to-speech and traditional pipelines, tool execution during calls, and agent behaviors that feel natural (listening cues, interruption handling, and consistent outcomes).
• Develop platform APIs and the admin experience: design and implement the configuration, publishing/versioning, and testing surfaces that let teams safely create, iterate, and operate voice agents.
• Knowledge + retrieval features: build/extend knowledge base and retrieval patterns (RAG) so agents can use customer context accurately and safely.
• Integrations and eventing: create clean integration layers, authentication patterns, and webhook/event systems that connect the agent platform to Aloware and external systems.
• Ship with an AI-enabled quality bar: use tools like Cursor/Claude Code to move faster while maintaining strong engineering standards—clean architecture, readable code, tests where it matters, and PR discipline (no “AI slop”).
  
  

What “Great” Looks Like in This Role

Within the first 30 – 60 days, you can ship production features that move us toward Retell/Vapi parity, then unlock tight Aloware integration, and finally deliver embeddable widgets + rollout tooling to migrate customers.

Requirements

• Strong production engineering experience (you’ve shipped systems that real customers depend on).
• Experience building with Python services and/or real-time, latency-sensitive systems.
• Comfortable with TypeScript + React/Next.js and full-stack ownership when needed.
• Familiarity with LLM systems and at least one of:
  • realtime/S2S voice models, or
  • STT/LLM/TTS pipelines, or
  • telephony/voice/streaming systems.
    
    

Nice to Have

• Realtime voice/telephony: SIP/WebRTC, call routing/transfers, contact-center workflows, latency-sensitive systems.
• LLM/agent building: tool/function calling, prompt/guardrails, evals and regression testing.
• Knowledge + retrieval (RAG): embeddings/vector search, ingestion from docs/URLs, relevance tuning.
• SaaS integrations: OAuth/JWT, webhooks, third-party APIs, reliable background jobs (retries/idempotency).

Team & Working Style

• Small team, high ownership, high standards.
• Fast iteration with AI tools — but every change must be explainable, testable, and maintainable.
• You’ll work directly with engineering leadership and product.
  

As part of this position, you are an engineer on-call on a rotational basis every two weeks following deployments. While emergencies don’t happen often, we do require an engineer on-call to maintain our quality of service.


With a global team spanning 15+ countries we value and celebrate diversity where every individual feels empowered to contribute their unique talents, ideas, and experiences, driving our global success

**Role open and available for Latam and Europe remote ONLY**