About the job

Can you see yourself revolutionising the Agentic AI industry? We are a multi-award-winning AI and SaaS provider based in Manchester, dedicated to boosting productivity and efficiency across our global customer base spanning five continents. As an ASR Data Scientist, you will be the architect of how our agents "hear" and understand speech.

This role offers a world-class research environment: you will find yourself in a team of like-minded research scientists who are supported by our specialist in-house Data Annotation team, allowing you to move beyond public datasets and curate bespoke, high-quality data for custom model training. While your primary focus is real-time and multilingual ASR, we foster a cross-functional culture where you can flex into Text-to-Speech projects, enabling you to develop a holistic understanding of the full AI Agent cycle.

Responsibilities

• Tackle state-of-the-art ASR challenges, including Voice Activity Detection (VAD), turn detection, and multilingual speech recognition.
• Conduct cutting-edge research on LLM-based speech recognition and
• audio-tokenisation, and push core conformer-based ASR architectures to their limit.
• Pair with junior team members to drive research goals and foster collaborative
• learning
• Contribute to original research, publish papers, and represent the company at global AI conferences.
• Maintain steady biweekly progression within our sprint-based research environment.
• Write concise technical documentation and research papers.

Requirements

• PhD in Computer Science, Speech & Signal Processing, or a related field.
• Alternatively, an MSc and 2+ years of commercial or academic research experience in Speech or LLMs.
• Strong foundational understanding of Neural Networks, Large Language Models, and Conformer/Transformer architectures.
• Expert-level proficiency in Python and PyTorch.
• Proven ability to drive independent research with minimal supervision.
• Strong oral and written communication skills.