Senior Machine Learning Engineer (Speech Synthesis)

Telnyx LLC; Kraków; On-site; PLN 350,000 - 550,000; Full time

Benefits offered by this job

Flexible work options

Unlimited personal time off

Location-based benefits

Job summary

Telnyx LLC in Kraków is building a next-generation multilingual TTS platform. You will define the stack, architecture, and best practices for training and deploying cutting-edge speech synthesis models that power our voice AI agents.

As a senior ML engineer, you will own the end-to-end training and inference pipelines, optimize for sub-100 ms latency, and push state-of-the-art techniques from research to production at global scale.

Qualifications

• Hands-on experience with neural TTS or related speech-synthesis systems.
• Proven ability to lead small teams and set technical direction.
• Strong background in production-grade ML systems for real-time speech.

Responsibilities

• Own the ML training and inference pipelines for multilingual speech synthesis.
• Design and implement scalable training architectures and deployment pipelines.
• Optimize for sub-100 ms latency in real-time speech generation.
• Evaluate emerging TTS techniques and bring them to production at scale.

Skills

Neural TTS

Speech Synthesis

Distributed Training

Multilingual Modeling

Low Latency

LLM-based Approaches

Job description

• As a Senior ML Engineer (Speech Synthesis), you’ll be a founding member of the team building Telnyx’s next-generation speech synthesis systems
• This is a greenfield opportunity — you’ll define the stack, architecture, and best practices for training and deploying state-of-the-art multilingual text-to-speech (TTS) models that power our voice AI agents
• You’ll build everything from distributed training pipelines to inference services that generate ultra-low-latency, lifelike voices across dozens of languages
• Your work will bridge research and production — shaping how millions of people experience real-time conversational AI
• Own the stack from day one: Design and implement the ML training and inference pipelines for multilingual speech synthesis
• Low-latency TTS: Engineer systems optimized for real-time, streaming speech generation with sub-100ms response targets
• Train cutting-edge models: Build and fine-tune multilingual TTS systems using modern architectures — including LLM-based, diffusion, and flow-matching approaches
• Massive-scale data processing: Develop pipelines for ingesting, aligning, and normalizing text, audio, and phonetic data across dozens of languages
• Experimentation at scale: Run distributed training across multi-node GPU clusters, tracking results and iterating quickly
• Cross-functional collaboration: Work with infrastructure and voice platform teams to deploy models that scale globally
• Research meets production: Evaluate emerging techniques (LLM-guided synthesis, zero/few-shot voice cloning, full-duplex modeling) and bring them to life in production-grade systems

Benefits

• Flexible, global team: Reaching our full potential means hiring people from all walks of life. Telnyx allows employees to work from home, and we offer in-person options for team members in Amsterdam, Chicago, Dublin and Warsaw
• Modern life, modern accommodations: Working in the business of innovation means offering support to our team to help them stay innovative. Telnyx offers unlimited personal time off to help employees recharge
• Benefits based on location: With team members across the globe, Telnyx offers competitive perks and benefits based on location. Employees in Ireland, the UK, and the US are eligible for healthcare benefits, and US employees also have access to 401(k) matching

Experience leading small teams and defining technical direction or team executablesProduction mindset: You build systems that run fast, stay stable, and are easy to maintainHands-on expertise with neural TTS, speech synthesis, or adjacent areas (ASR, voice cloning, multilingual modeling)You’ve obsessed over one or two hard problems, whether it’s building multilingual TTS from noisy data, teaching LLMs to speak, designing self-supervised audio encoders, or making diffusion models run in real timeExperience with LLM-based approaches to speech synthesis or prosody control6+ years of experience in machine learning or speech systems engineering