Machine Learning Engineer, TTS

This is a fully remote position, open to applicants in Serbia.

📋 Description

• Expand TTS capabilities to new languages through data collection and curation, text normalization/G2P, and model adaptation.

• Transition new TTS architectures from research phases to production environments.

• Enhance model robustness, reduce latency, and improve quality while working closely with MLOps during deployment.

• Advance existing models through post-training methods, data curation, architectural upgrades, and innovative training strategies.

• Develop efficient audio data pipelines, covering speaker recording and preprocessing to model training.

• Create comprehensive benchmarks and evaluations to assess whether new model iterations exhibit improvements and the reasons behind them.

• Become a member of a compact TTS team consisting of 2–3 ML engineers, with the backing of a dedicated MLOps team and a data engineer.

⛳️ Requirements

• A minimum of 3 years of experience in speech technology, specifically with hands-on TTS expertise.

• Practical experience with contemporary LLM-based TTS methodologies (e.g., VALL-E, CosyVoice, XTTS).

• Experience in making TTS models production-ready, focusing on stability, robustness, and inference efficiency at the model level.

• Proficiency in deep learning frameworks, particularly PyTorch.

• A solid understanding of modern ML principles, including Transformers, diffusion/flow matching, neural audio codecs, and vocoders.

• Familiarity with streaming TTS constraints (latency, chunked/incremental generation) and their influence on model architecture decisions.

• Knowledge of audio signal processing along with the linguistic aspects of TTS (phonetics, text normalization).

• Basic experience in speech data processing, including segmentation, alignment, and quality filtering.

• Understanding of TTS evaluation techniques (MOS, SBS).

• Strong skills in Python and a solid foundation in software engineering best practices.

• Demonstrated ability to independently investigate new concepts and enhance existing ML systems.

• A BS degree or higher in Computer Science or a related field, or equivalent work experience.

• Proficiency in English at a B2 level or higher.

• Experience with low-resource languages, including dataset creation and augmentation.

• Background in designing recording scripts and specifications for TTS voice datasets.

• Experience in MLOps/DataOps, such as MLflow, DVC, or comparable experiment tracking tools.

• Skills in C++.

• Publications in the field of speech or ML.

🏝️ Benefits

• The team has developed award-winning AI products for technology companies, including devices, voice assistants, and products that are actively utilized in the market.

• Access to a cutting-edge tech stack that includes Speech Technologies, NLP, Generative AI (LLMs, diffusion models), and a voice-first agent architecture emphasizing privacy and on-premises deployment.

• A high engineering standard with genuine ownership — the team prioritizes what effectively functions in production over mere visual appeal in demos, allowing you to witness the direct impact of your contributions.

• Rapid career advancement opportunities — a senior-heavy team coupled with a high volume of real-world challenges enables faster growth compared to other environments.

• A startup atmosphere combined with enterprise-level stability — working with real clients, generating actual revenue, and experiencing minimal bureaucracy.

• Fully remote work opportunities across Europe.

• 21 vacation days in addition to public holidays and 5 sick days.

• Private English lessons provided through Preply.