Background

Curriculum Vitae

Education

4.0 GPA (94.5% average)
Awarded a $10,000 NSERC Responsible AI Group Grant for research in interpretable AI under Dr. Robin Cohen.

Accelerated automotive weld inspection by 300×, reducing analysis time from 5 minutes to 1 second, by developing a U-Net segmentation model achieving 0.90 IoU on a held-out test set for non-destructive weld evaluation at Centerline.
Developed a Python/scikit-learn based pipeline to automate augmentation, preprocessing, and quality evaluation for 15,000+ weld images, reducing multi-hour data workflows to minutes.
Resolved 3 critical model issues by building an inspection toolkit to visualize ultrasonic signals alongside segmentation outputs and designing domain-specific evaluation metrics for faster experimental iteration.

Cut multilingual model evaluation time by 50%, from 20 minutes to 10 minutes, by engineering Hydra-configured Stopes pipelines scaled across 128+ GPU nodes on Meta's HPC cluster with SLURM.
Benchmarked 25 SeamlessM4T speech-translation models across 101 languages and 12 datasets per MLOps iteration, by systematically distributing evaluations across massive horizontal scale.
Integrated the evaluation pipeline into Meta's open-source Stopes framework, enabling continued use by Meta AI engineers.

Expanded the LADy benchmark from 2 to 10+ heterogeneous review datasets, increasing evaluation coverage 5×.
Benchmarked RePair on literature-derived baselines by designing evaluation modules and conducting NLP experiments.

Led 10+ student developers shipping a Next.js platform and AI tools for a community of 2,500+ students.
Built a RAG student assistant over 1,000+ documents using Qdrant retrieval and Cohere Rerank, handling 50+ daily queries and achieving 0.91 nDCG@10.
Reduced manual moderation work by developing a BERT-based classification system that achieved a 0.88 F1 score.