Victory Chianumba
Research engineer and MSc AI graduate working on mechanistic interpretability, adversarial robustness, and research tooling
Hire this person
Send a job offer directly to this candidate
About
Research engineer and MSc AI graduate working on mechanistic interpretability, adversarial robustness, and research tooling. Recent findings include a class of low-pass adversarial attacks that flip EEG classifier predictions while leaving Integrated Gradients explanations effectively unchanged (0% detection rate where standard attacks are caught 87–97% of the time), and a case study showing activation-magnitude feature ranking — the LLM-SAE default — surfaces persistent state-trackers over event detectors when imported to a world-model substrate.
Experience
Data Collection Officer at Appen (2023-01 – 2024-12)
Freelance Composer / Audio Engineer at Self-employed (2020-01 – 2023-12)
Education
MSc in Artificial Intelligence for Engineering and Design – University of Bath (2024-09 – 2026-01)
BSc (Honours) in Computer Science – University of Essex (2016-10 – 2020-07)
Year abroad in Machine learning and functional programming – University of Wyoming (2018-01 – 2019-12)
