Overview

Work within our clients machine learning team to deploy and optimize models for applications like low-latency speech recognition and large language models (LLMs). Initial focus will be on improving our clients speech recognition model’s training pipeline on multi-GPU systems to boost performance and quality.

Hybrid working, 3 days onsite and 2 days WFH.

No Sponsorship, must have the right to work in the UK.

Responsibilities:

• Train and deploy state-of-the-art ML models.
• Apply optimization techniques (distillation, pruning, quantization).
• Enhance speech models with features such as diarization, multilingual support, and keyword boosting.
• Optimize models for low-latency inference on accelerators.
• Improve training workflows and GPU utilization.
• Use data augmentation to improve performance.
• Stay updated on ML research to guide strategy.

Requirements:

• Master’s or PhD in a relevant field with strong ML foundations.
• Training ML models for production use.
• PyTorch or TensorFlow.
• Handling large datasets (multi-terabyte).
• Familiarity with Linux, version control, and CI/CD systems.
• Knowledge of model compression (e.g., reduced precision).