MARIANGLEN LOUIS

Conducting applied research on time series trend prediction and root-cause analysis for power industry datasets under Dr. Junsong Yuan (IEEE Fellow, Director — Visual Computing Lab, UB CSE), targeting submission to a reputable conference or journal.

Designing and benchmarking forecasting pipelines to model temporal patterns in power consumption data, evaluating statistical and ML baselines for predictive accuracy and interpretability of identified trend drivers.

Collaborating with PhD researcher Ziqing Zhang to structure the data analysis workflow, define evaluation criteria, and build reproducible experiment pipelines for academic publication.

Engineered a real-time Computer Vision quality control system using YOLOv8 Nano to inspect 30 million tickets daily, achieving 88.1% mAP and sub-100ms inference times to eliminate high-speed production bottlenecks.

Developed a defect analysis pipeline evaluating pixel color intensity and bounding box dimensions, capturing 86.67% of critical micro-defects while maintaining 88.45% precision to ensure no good material was wasted.

Built a predictive maintenance module tracking dimensional drift and color deviation over time, establishing specific warning thresholds (e.g., Delta E > 15) to proactively alert operators to perform maintenance before production failures occurred.

Pipeline Engineering: Architected Python-based validation frameworks to automate data reliability for large-scale Snowflake migrations, ensuring integrity for downstream ML analytics.

Technical Leadership: Directed a team in requirement analysis and SQL development, establishing peer-review processes to guarantee high-fidelity data deliverables.

AI Guardrails: Established Ground Truth benchmarks to reconcile backend records with BI metrics, preventing decision-making hallucinations in automated layers.