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.