Overview
As a Data Scientist on the Text-to-Speech (TTS) team, you will shape the voice of ConnexAI, bringing conversational intelligence to life through natural, expressive, and humanlike speech generation. You’ll research, design, and train SOTA TTS models that combine neural acoustic modelling, LLMs, and multimodal learning to produce voices that sound truly alive. Your work will push expressive prosody, emotional control, speaker adaptation, and contextual intonation to ensure our AI voices sound real and feel real. This is a pivotal moment for ConnexAI, as we expand our TTS capabilities and redefine the standard for expressive conversational speech.
This is your opportunity to pioneer the next generation of voice synthesis technology, used daily by millions of users and enterprise clients worldwide.
Core Responsibilities
- Research, design, and train SOTA LLM-based and neural TTS models, combining diffusion, flow-matching, or autoregressive architectures for ultra-natural speech generation.
- Explore and implement expressive prosody modelling, emotion-conditioned synthesis, and speaker cloning for personalised voice experiences.
- Develop text–audio alignment and context-aware TTS pipelines that adapt to semantic intent, dialogue context, and emotional tone.
- Work with large-scale speech and text corpora, managing data cleaning, augmentation, and multi-speaker dataset preparation for model training.
- Collaborate with speech engineers and linguists to integrate linguistic, phonetic, and prosodic features into training workflows.
- Optimise models for real-time inference, ensuring low latency and high perceptual quality for live conversational systems.
- Support deployment of models across ConnexAI’s AI Voice and Conversational AI platforms, ensuring seamless integration with ASR, NLP, and dialogue systems.
- Stay current with cutting-edge research in neural TTS, speech synthesis, and generative modelling, rapidly translating new ideas into production-ready improvements.
Requirements
- MSc or PhD in Speech & Signal Processing, Computational Linguistics, Computer Science, Data Science, or a related field.
- At least two years of hands-on commercial or academic experience in designing, training, and evaluating neural models for speech or language generation.
- Strong understanding of large language models (LLMs), transformers, and sequence-to-sequence learning.
- Advanced proficiency in Python and PyTorch, with practical experience in model training and optimisation at scale.
- Solid software engineering practices: Git, unit testing, and CI/CD pipelines.
- Proven ability to conduct independent research, troubleshoot complex modelling issues, and deliver high-quality results with minimal supervision.
- Passion for voice synthesis, expressive modelling, and creating AI voices that connect with people naturally.
Interview Process
- 30-minute video call with the team lead
- Take-home technical exercise
- 90-minute face-to-face interview
About ConnexAI
ConnexAI is an award-winning Conversational AI platform. Designed by an elite engineering team, ConnexAI’s technology enables organisations to maximise profitability, increase revenue, and take productivity to new levels.
ConnexAI provides cutting-edge, enterprise-grade AI applications, including AI Agent, AI Guru, AI Analytics, ASR, AI Voice, and AI Quality.
We value growth both for our products and our people. As we scale, there will be clear opportunities to progress into senior data science, leadership, or principal research roles. Our high retention rate reflects our inclusive, supportive, and empowering environment.
Role Details
- Seniority level: Associate
- Employment type: Full-time
- Job function: Research and IT
- Industries: IT Services and IT Consulting, Software Development, and Research Services
Note: This description focuses on the position and removing unrelated job board noise. Referrals and other job postings mentioned in the original listing are not part of the role description.
