Senior Data Scientist

Connect; City Of London; On-site; GBP 90,000 - 150,000; Full time

Job summary

Connect is seeking a highly experienced Senior Data Scientist to lead research, development, and scaling of conversational AI and multimodal systems with a focus on Speech Processing. You will own end-to-end ASR model lifecycles—from data pipelines to deployment—and drive performance in demanding environments.

You will architect scalable labeling workflows, optimise training via quantisation and distillation, and guide decisions on custom vs foundation models.

Qualifications

• Master's or Ph.D. in Speech Science, AI, Computational Linguistics, Mathematics, or related quantitative field.
• 5+ years in ML/DL/Data Science, with at least 2 years in Generative AI, NLP, or Speech.
• Deep theoretical and practical knowledge of acoustic modelling and ASR architectures (Whisper, Conformer, wav2vec, Nemotron).
• Experience designing data curation strategies and managing audio/text annotation workflows.
• Proficient in Python and PyTorch.

Responsibilities

• Lead the end-to-end lifecycle of core ASR models, from data pipelines to training and deployment.
• Build and optimise custom ASR models for high-accuracy transcription in noisy environments.
• Architect scalable data pipelines for unstructured audio and text, with QA rubrics and human-in-the-loop labelling.
• Guide model evaluation and production inference optimisation, including quantisation and distillation.
• Evaluate emerging architectures and decide when to build or fine-tune Foundations models.
• Collaborate with ML Engineers, Backend Developers, and PMs to meet latency and throughput goals.
• Mentor junior data scientists and engineers, promoting rigorous experimentation and robust engineering.

Skills

Python

PyTorch

Education

Master's or PhD in Speech Science or related quantitative field

Tools

Whisper

Conformer

wav2vec

TensorRT

Job description

About the Role

We are seeking a highly experienced Senior Data Scientist to lead the research, development, and scaling of our next-generation conversational AI and multimodal systems. In this role, you will focus on Speech Processing. You will own the end-to-end lifecycle of our core models — from designing massive data pipelines and human-in-the-loop labelling systems, to training, fine-tuning, and deploying state-of-the‑art Automatic Speech Recognition (ASR) models.

Key Responsibilities

• Speech technology (ASR): Build, train, and optimise custom ASR models for high-accuracy transcription in noisy environments.
• Data processing and labelling pipelines: Architect scalable data pipelines for ingesting, cleaning, and processing large volumes of unstructured audio and text data. Design strict quality assurance rubrics and manage human-in-the-loop labelling workflows to ensure high-fidelity training datasets.
• End-to-end model lifecycle: Lead the full iterative lifecycle, including data curation, baseline modelling, hyperparameter optimisation, evaluation, and production inference optimisation using techniques such as quantisation, distillation, and TensorRT.
• Architectural leadership: Evaluate emerging architectures, research papers, and open-source models. Make strategic decisions on when to build custom models from scratch versus fine-tuning existing foundational models.
• Cross-functional collaboration: Partner with ML Engineers, Backend Developers, and Product Managers to integrate models into user-facing applications with stringent latency and throughput requirements.
• Mentorship: Guide and mentor junior data scientists and engineers, fostering a culture of rigorous scientific experimentation and robust engineering practices.

Required Qualifications

• Education: Master's or Ph.D. in Speech Science, Artificial Intelligence, Computational Linguistics, Mathematics, or a highly related quantitative field.
• Experience: 5+ years of industry experience in Machine Learning, Deep Learning, or Data Science, with at least 2 years specifically focused on Generative AI, NLP, or Speech domains.
• Speech processing: Deep theoretical and practical understanding of acoustic modelling, ASR architectures such as Whisper, Conformer, wav2vec, Nemotron
• Data engineering for AI: Extensive experience designing data curation strategies, managing data annotation projects, and using tools for audio processing and text normalisation.
• Programming and frameworks: Expert-level proficiency in Python and PyTorch.

Preferred Qualifications

• Experience deploying models on edge devices or optimising ML models for low-compute environments.
• Published research in tier-1 conferences such as ACL, Interspeech, ICASSP, or NeurIPS.
• Experience with multimodal model architectures, including audio-text native models.
• Familiarity with cloud ML ecosystems such as AWS SageMaker, GCP Vertex AI, or Azure ML.