Terry Oko-odion

Co-evaluated the clinical safety of multimodal LLMs across text and speech modalities, quantifying audio penalty performance drops in low-resource languages. Developed shift-aware safety metrics using PyTorch and co-led a statistical audit of the XNLI benchmark across 11 African languages.

Built and maintained the Briscoe website AI chat feature, automating first-level customer responses and reducing repetitive enquiry handling by an estimated 30-40%. Architected and managed the company's internal IT ticketing infrastructure, optimizing SLA response times and troubleshooting critical system bottlenecks.

Engineered a balanced data distribution strategy for the Goldberg microbial signature dataset to mitigate spurious temporal correlations. Restructured clinical metadata for 233 individuals and collaborated with domain experts to establish clinically rigorous evaluation metrics.

Designed and built RESTful APIs using Ruby on Rails to power core transit logging features. Wrote and managed database migrations and schema changes using PostgreSQL, and developed backend features for a microservice architecture.

Developed and optimized a Convolutional Neural Network (CNN) using PyTorch to classify food items from images and estimate calorie content. Engineered a hand gesture recognition pipeline and implemented SVM and K-means clustering models for image classification and customer segmentation.

Conducted User Acceptance Testing (UAT) and diagnostic device benchmarking on unreleased versions of the MyMTN mobile application. Partnered with external security assessors to scan and audit 1,000+ enterprise network nodes, and administered threat monitoring pipelines and security alerts.

Completed a Bachelor of Science in Computer Science in July 2024.