4 hours ago
Ho Chi Minh City, VietnamMid Level

Responsibilities

  • Own real-time orchestration of the ASR-to-LLM-to-TTS cascade, including streaming, state management, and timing.
  • Build and maintain speech understanding, speech generation, voice activity detection, end-of-turn detection, barge-in, and interruption handling.
  • Reduce end-to-end latency from incoming audio to first audio output.
  • Integrate and switch between ASR/TTS vendors and self-hosted speech models through tested interfaces.
  • Deploy and scale self-hosted speech models on GPUs using batching, streaming inference, and cost-per-call optimization.
  • Own monitoring, alerting, load testing, reliability, and incident response for the interaction layer with Datadog.
  • Work with Data Scientists to productionize model improvements through A/B tests and staged rollouts.
  • Collaborate with data science, product, and engineering teams on speech, LLM-script, telephony, and inbound-routing interfaces.

Requirements

  • At least 3 years of backend or ML engineering experience with production real-time or streaming systems.
  • Strong Python skills and experience with asynchronous or concurrent programming.
  • Experience with WebSocket- or WebRTC-based streaming.
  • Hands-on experience with real-time voice frameworks such as LiveKit or Pipecat, or equivalent media pipelines.
  • Experience integrating ASR/TTS APIs and working with audio formats, codecs, buffering, and jitter.
  • Experience across modeling, engineering, deployment, and production monitoring.
  • Solid engineering practices involving testing, CI/CD, containers, cloud deployment, and observability.
  • Preferred experience serving speech or LLM models on GPUs with tools such as vLLM, Triton, or TensorRT.
  • Familiarity with SIP and telephony concepts is preferred.
  • Experience with speech-to-speech models or cascade-versus-S2S architecture trade-offs is preferred.
  • Vietnamese, Tagalog, or US English voice product experience is preferred.

Benefits

  • Competitive compensation package with a 13th-month salary and performance bonuses.
  • Comprehensive healthcare coverage for employees and dependents.
  • Generous annual and sick leave policies with flexible work hours.
  • Central District 1 office location near a future metro station.
  • On-site lunch with multiple meal options, including vegetarian choices.
  • Grab for work allowance and fully equipped workstations.
  • Team-building activities, sponsored sports clubs, and weekly Thursday happy hour.
  • Unlimited coffee, tea, snacks, and fruit.
  • Opportunity to work on social-impact products that expand credit access in emerging markets.
  • The posting describes a central office location but does not explicitly state a remote or hybrid arrangement.

Tech Stack

DatadogPython
Trusting Social & Kompato AI

About Trusting Social & Kompato AI

201-500 employees
Contact me