Sergius Justus C. Nyah

Machine Learning Researcher & Software Engineer

Gmail

01About

I'm a Machine Learning Researcher and Software Engineer. I work on causal and multilingual NLP, LLMs, medical AI, and geometric deep learning. I'm currently an ML Researcher at the University of Toronto's Jinesis AI Lab. I hold a B.S. in Computer Science from the University of Buea. My M.Sc. starts in November 2026.

My research began at MIT CSAIL, through the MIT Summer Geometry Initiative. I was 1 of 32 global fellows, and one of two from Africa. I worked as a Research Fellow under Professor Justin Solomon's Geometric Data Group, mentored by Dr. Karthik Gopinath of Harvard Medical School. That work became my first-author paper at IEEE EMBC 2026.

I later researched multilingual LLM robustness at the Fatima Institute for Global AI Research, with the Stanford NLP Group and Georgia Tech. I co-authored "To Lie or Not to Lie?", accepted to ACL 2026 with a Senior Area Chair Highlight Award. I also spent a summer at EPFL's LiGHT Lab, building survival models for Ebola Virus Disease.

On the engineering side, I've shipped production ML and full-stack systems at Engineering for Change and Hometeam Ventures. I kept telecom infrastructure running at Huawei. I build retrieval pipelines, research platforms, and monitoring systems that real teams use daily.

Outside research, I'm a Christian. I play piano and bass guitar, and led music ministry at CMFI Molyko in Buea. I support FC Barcelona.

02Selected Experience

View full work history →

Machine Learning Researcher

Jinesis AI Lab, University of Toronto·September 2026 – Present

Research Intern

Intelligent Operations Engineer

Huawei·August 2025 – February 2026

04Featured Research

ACL 2026 · Main ConferencearXiv:2604.06552

To Lie or Not to Lie? Investigating the Biased Spread of Global Lies by LLMs

LLMs spread misinformation unevenly. Propagation is higher in lower-resource languages and regions with lower Human Development Index scores. We introduce GlobalLies, a multilingual benchmark spanning 195 countries, 8 languages, and 6,867 entities. It exposes systematic gaps in AI safety across global contexts.

195
Countries
8
Languages
6,867
Entities
NLPLLMsMisinformationMultilingualAI SafetyFairness

Z. Khan · M. Dogan · I. Okoh · P. Sadeghi · S. Shrestha · S. J. Nyah · M. O. Mokhiamar · M. J. Ryan · T. Naous

05Featured Writing

Resilience5 min read

The Plane Tickets I Never Bought

Four visa rejections, lessons learned, and the God who never left, across a COSYNE talk, an EPFL internship, and an EMBC 2026 travel grant.

Jul 23, 2026