CV
Manx-speaking PhD researcher at the University of Sheffield working on data-efficient speech technology for under-resourced and critically endangered languages. Background in linguistics, with a prior career in financial-crime analysis.
Education
PhD, Computer Science
University of Sheffield · 2024–present
Doctoral research in data-efficient speech technology for endangered languages, in the Speech and Hearing Research Group (SpandH). Topics span automatic speech recognition, text-to-speech, language identification, language modelling, and statistical machine learning.
MA, Linguistics
University of Manchester · 2022–2024 · Distinction
Computational linguistics, sociolinguistics, semantics, and phonology. Dissertation: Evaluating Neural Approaches to Nuanced Spanish–English Translations: A Challenge Set Approach.
BA, Modern Languages
University of Chester · 2015–2019 · First Division (2:1)
Spanish, French, and Mandarin, with translation techniques and a year-abroad placement.
Research
Data-efficient ASR for critically endangered languages
University of Sheffield · 2024–present
Built the first utterance-level speech corpus for Manx Gaelic with a Kaldi forced-alignment pipeline over 300+ hours of collected recordings, and used it to train the world's first ASR system for the language. Reproduced the approach for Cornish, Hawaiian, Jeju, and Mohawk, showing competitive recognition at a fraction of the compute of large multilingual models, and released the corpora and tooling as open resources. Accepted to Interspeech 2026; current work targets read↔spontaneous domain mismatch and unsupervised ASR.
Experience
Gaelg AI — creator & maintainer
2024–present
Built and maintain gaelgai.im, a public platform of AI tools for Manx Gaelic (speech synthesis, recognition, translation, and audio timestamping). Responsible for model training, full-stack web development, server administration, and community engagement.
Teaching assistant — COM4511/6511 Speech Processing
University of Sheffield
Marking and feedback across coursework on MFCC feature extraction, keyword spotting, voice activity detection, K-means clustering, and Gaussian mixture models.
Data annotation & evaluation — contract
Adversarial prompt construction and evaluation-rubric design for AI search agents, and response evaluation for meeting-transcript summarisation.
Earlier career
Anti-Money Laundering (AML) Investigator — The Co-operative Bank
2022–2024 · Manchester
Analysed transaction data to identify suspicious activity and raise SARs, reported on the efficacy of automated transaction-monitoring rules to stakeholders, and liaised with law enforcement.
Fraud Analyst — Suits Me
2021–2022 · Knutsford
Assessed transaction-monitoring efficacy, raised SARs with the National Crime Agency, managed fraud data and stakeholder reporting, and trained staff.
Latin America Cell Associate — Quilter International
2020–2021 · Isle of Man
Customer due diligence under anti-money-laundering regulations, authorisation of large transfers, and Spanish-language client security calls.
Technical skills
Speech & ML: Python, PyTorch, Kaldi, Fairseq, ESPnet, SpeechBrain, Hugging Face Transformers, scikit-learn, pandas, NumPy; forced alignment, acoustic and language modelling (HMM-based and end-to-end), MFCC/filterbank features, KenLM/SRILM, Grad-TTS, DiffVC.
Tools & infrastructure: Linux, HPC (SLURM), Git, Docker, Conda, LaTeX; NGINX/SSL server administration and full-stack web development.
Data: SQL, SAS, Tableau, data curation and annotation.
Certifications
- Deep Learning Specialisation — DeepLearning.AI (2023)
- Python for Everybody — University of Michigan (2022)
- Data Analytics — Google (2022)
Languages
English (native) · Spanish (professional) · French (professional) · Manx (professional) · Mandarin (basic)
Publications
-
Bootstrapping Endangered Language ASR with Short-Form Corpora
Christopher Bartley, Anton Ragni
Interspeech 2026 -
How I Built ASR for Endangered Languages with a Spoken Dictionary
Christopher Bartley, Anton Ragni
Preprint · 2025 · arXiv
Contact
csjbartley1@sheffield.ac.uk · GitHub · LinkedIn · ORCID · Google Scholar