Research Scientist, Text-to-Speech

Stack

Responsibilities

• ▹Research and train fast, quality SOTA TTS models for realistic and emotional voice generation
• ▹Experiment with different architectures and data to improve quality and speed, and ship the best results to production
• ▹Stay current with research and propose new ideas / improvements
• ▹Collaborate closely with a team of 3 TTS researchers, supported by engineering and hardware

Requirements

• ▹Experience training text-to-speech / voice cloning models
• ▹Solid knowledge of transformers, diffusion models, GANs
• ▹Understanding of human speech and audio processing (sampling, spectrograms, vocoders)
• ▹Proficiency in Python and key libraries (PyTorch, Hugging Face Transformers)
• ▹Ability to keep up with research, understand papers, and implement approaches; strong ML fundamentals and critical thinking

Nice to have

• ▹Familiarity with modern speech synthesis models (GPT-based, flow matching, e.g. Vevo, StyleTTS, IndexTTS, Maskgct)
• ▹Contributions to open-source AI tools or research publications in speech processing
• ▹Familiarity with AWS or similar clusters

Soft skills

About the company

Valka is a spin-off from the Realms Group (parent company of Oddin.gg) building an interactive, generative-AI digital content platform where virtual characters respond dynamically to users' voice, text, and gestures in real time.