Postdoctoral Researcher: Multilingual LLMs & Open-Source Benchmarks

UAG; Oxford; Hybrid; GBP 39,000 - 42,000; Full time

Job summary

UAG in Oxford is offering a fixed-term 12-month role to deliver the BodleianLLM project: curating multilingual corpora, continual pre-training and fine-tuning on the University's infrastructure, and developing benchmarks for humanities tasks.

You will publish model weights, code and benchmarks as open-source resources, contribute to funding bids, collaborate on publications and workshops, and report to the PI with day-to-day supervision from DiSc.

Qualifications

• PhD or substantial ML research/industry experience.
• Strong Python and software engineering skills.
• Proven publication record.
• Good communication and organisational abilities.
• Desirable: experience with cultural heritage data, historical languages, HPC and MLOps.

Responsibilities

• Curate, clean, and document a multilingual humanities training corpus.
• Pre-train and fine-tune open-weight language models on Oxford infrastructure.
• Design and validate evaluation benchmarks for humanities tasks.
• Publish model weights, code, and benchmarks as open-source.
• Contribute to research roadmap and project workshops.

Skills

Python

NLP

Deep Learning

Software Engineering

Education

PhD / DPhil in ML or related field

Tools

Linux HPC

Git

Docker

Job description

UAG in Oxford is offering a fixed-term 12-month role to deliver the BodleianLLM project: curating multilingual corpora, continual pre-training and fine-tuning on the University's infrastructure, and developing benchmarks for humanities tasks.

You will publish model weights, code and benchmarks as open-source resources, contribute to funding bids, collaborate on publications and workshops, and report to the PI with day-to-day supervision from DiSc.