Manager Data Science – LLM Training and Team Lead

About the role

A Manager Data Science at LexisNexis Legal & Professional is an emerging subject‑matter expert who leads a team of junior data scientists. The role combines technical leadership with mentorship, ensuring best‑practice adoption while defining innovative approaches to business problems.

Key responsibilities

• Lead design and execution of LLM training and fine‑tuning projects, including model selection, strategy, experimentation, and evaluation.
• Oversee creation of high‑quality training datasets: collection, cleaning, deduplication, annotation, and validation.
• Develop and optimise supervised and parameter‑efficient fine‑tuning workflows, applying preference‑optimisation methods where appropriate.
• Establish evaluation frameworks for factual accuracy, instruction following, domain relevance, safety and business‑specific performance.
• Diagnose training issues, improve model quality, training stability, GPU utilisation and computational efficiency.
• Mentor and manage data scientists, review technical work and enforce reproducible development practices.
• Partner with product, engineering and domain experts to define requirements, support model deployment and monitoring.
• Manage project priorities, timelines, compute resources and communicate results and trade‑offs to stakeholders.

Required profile

• Strong understanding of transformer architectures, attention mechanisms, tokenisation and LLM training fundamentals.
• Proficient in Python and PyTorch, with hands‑on experience using Hugging Face Transformers and Datasets.
• Demonstrated ability to implement supervised fine‑tuning, hyper‑parameter tuning and checkpoint selection.
• Experience with parameter‑efficient fine‑tuning techniques such as LoRA or QLoRA.
• Skilled in GPU‑accelerated and distributed training using frameworks like PyTorch FSDP or DeepSpeed.
• Ability to design reliable benchmarks, conduct human evaluations and troubleshoot training instability.

Required skills

• Python
• PyTorch
• Hugging Face Transformers & Datasets
• LoRA / QLoRA
• GPU and distributed training (FSDP, DeepSpeed)
• Experiment tracking and versioning

What we offer

• Healthy work‑life balance with flexible, remote working hours.
• Wellbeing initiatives, shared parental leave, study assistance and sabbaticals.
• Annual incentive bonus and a base salary range of $115,400 – $192,300 per year.