Coding-agent evaluation | Low-resource NLP | Responsible AI
Oyinkansola Onwuchekwa
I specialise in coding-agent evaluation, low-resource NLP, and responsible AI systems. I build tested Python tools, trace datasets, monitor harnesses, and evidence-grounded workflows for research and decision-making.
Research
PhD in Data Science and AI
Thesis research explores cross-cultural musical elements, emotional expression, and genre characteristics in contemporary global music lyrics using NLP and deep learning. Awarded a fully funded Faculty of Science and Engineering PhD Scholarship in Data Science.
Technical focus
I build tools that make agent behaviour auditable: labelled traces, intervention points, evidence spans, safe counterfactuals, and reproducible monitor tests.
Coding-Agent Evaluation
Building trace-level datasets and monitor experiments for coding-agent behaviour, with attention to evidence spans, intervention points, safe counterfactuals, and testable failure modes.
Low-Resource Emotion Recognition
Developing multilingual emotion recognition pipelines for African languages, with attention to class imbalance, code-mixed text, tokenisation, cross-lingual embedding alignment, and responsible model evaluation.
Responsible Cultural AI
Contributing to British Academy ODA and UNESCO-linked work on AI literacy, intangible cultural heritage, data governance, digital visibility, and community-led frameworks for cultural data use.
Publications and Presentations
Co-author and presenter on locally responsible AI frameworks for cultural heritage governance, with additional work in persuasion detection, multilingual NLP, invited talks, and public engagement.
Experience
Graduate Teaching Assistant and PhD Researcher
Develop and evaluate machine learning, NLP, and deep learning pipelines for multilingual emotion modelling. Teach machine learning, NLP, and deep learning laboratory sessions to 150+ MSc students, supporting Python, model development, and applied AI workflows.
AI Research Engineer
Designed components of a LangGraph-based multi-agent AI system for systematic literature review workflows, including protocol generation, screening, evidence synthesis, auditability, and reproducible structured outputs.
Research Contributor
Developed and refined responsible AI, digital literacy, and cultural data governance materials for intangible cultural heritage practitioners. Built Python and Hugging Face data processing workflows and contributed to facilitator-ready training materials for international partners.
Data Analyst and Social Researcher
Analysed programme and operational datasets with Python to evaluate food redistribution impact, improve reporting quality, and support evidence-based service decisions. Awarded Outstanding Achievement Intern Award.
Projects
Coding-agent safety and evaluation
Auditable traces and monitor tests.
Work focused on where coding agents fail, what evidence a monitor should inspect, and where intervention helps.
Coding Agent Failure Atlas
Labelled trace dataset
A synthetic atlas of coding-agent failure traces for monitor research, with evidence spans, intervention points, and safer counterfactuals.
Coding Agent Monitor Lab
Monitor evaluation harness
A small research lab for testing whether monitors can catch risky coding-agent traces using structured evidence rather than broad completion scores.
Low-resource NLP tooling
Language, emotion, and evaluation workflows.
Public work for multilingual NLP research, especially African language processing, code-mixed text, and emotion-aware evaluation.
Low-Resource NLP Toolkit
Released Python package
Utilities for African language pre-processing, emotion-label mapping, evaluation, language routing, and code-switching audits.
Dimensional ABSA Research
Multilingual sentiment research
Research code for predicting how people feel about specific aspects in text, using valence-arousal scores for finer-grained sentiment beyond positive or negative labels.
Service and Speaking
Academic and Professional Service
- UKRI Member, EPSRC and NERC Peer Review Colleges
- Reviewer, International Conference on Learning Representations
- Reviewer, The Deep Learning Indaba
- Associate Fellow, Advance HE
- Professional Member, BCS - The Chartered Institute for IT
Selected Talks and Engagements
- Panel Speaker, 6th European Chatbot and Conversational AI Summit
- Co-author and Presenter, Digital Humanities 2025, Lisbon
- Invited Speaker, PyCon Lithuania 2025
- Lead Organiser and Speaker, Towards Transparent and Responsible AI Conference
- Invited Speaker, Pint of Science UK, Hull
- Featured Contributor, BBC News Interview for National AI Day
Community and Mentoring
Volunteer academic support tutor with IntoUniversity Hull East, research mentor through University of Hull and Nuffield Research Placements, founder of Tech Help Group, and community contributor with Hull Afro-Caribbean Association.
Skills
AI and NLP
Multilingual NLP, low-resource language processing, emotion classification, LLM fine-tuning, retrieval-augmented generation, coding-agent evaluation, agent failure analysis, bias and fairness metrics, and model evaluation.
Data and Engineering
Data pipelines with pandas and NumPy, scikit-learn, Bash, trace schemas, structured outputs, reproducible documentation, Python packaging, and experiment-ready tooling.
Languages
English and Yoruba fluency, intermediate Spanish, and basic French, with research interests in multilingual and cross-cultural language technologies.
Education
University of Hull: PhD in Data Science and AI, 2023 - present; MSc Artificial Intelligence and Data Science, Distinction, 2022 - 2023.
University of Lagos: BSc Geophysics, Second Class Upper, 2015 - 2020.