• In-house LLM development: contribute to model selection, post-training, and release gates for domain-specific models supporting legal research workflows.
• Post-training data creation: design instruction-tuning and preference datasets using rubric-driven synthetic generation with human-in-the-loop QA.
• Auto-evaluation pipelines: build evaluation harnesses that run on every model drop, with task suites, reproducible seeds for stable comparisons.
• LLM-as-a-judge: implement multi-criteria rubric graders and multi-agent debate evaluation to reduce single-judge bias and improve reliability.
• Agent evaluation: assess tool-using agents with metrics for task success, tool-call accuracy, latency, and failure recovery in sandboxed environments.
• Cross-functional collaboration: partner with research, product, and legal subject-matter experts to translate evaluation results into model release criteria and product-ready guidance.

• Built an enterprise propensity-to-purchase model combining binary classification for propensity scores with regression for ACV prediction, improving lead prioritization and sales targeting.
• Performed extensive feature engineering and exploratory data analysis across firmographic, demographic, and engagement data sources.
• Partnered with cross-functional stakeholders to scope requirements, communicate results, and iterate on modeling choices to maximize business impact.

The Human Machine Lab at Ontario Tech University is an interdisciplinary research group focused on designing computer systems around human needs and capabilities, with projects spanning human–computer interaction, usable security, privacy, and artificial intelligence.

Advised by Dr. Miguel Vargas Martin on applied machine learning research:

• Published multiple peer-reviewed papers on applying machine learning techniques to authentication systems (see Google Scholar).
• Designed GAN-based password guessing models that outperformed prior benchmarks by 83%, and developed a GPT-3-based honeyword generation technique to accelerate password-breach detection.
• Conducted systematic literature reviews on machine learning applications in computer security.
• Proposed novel approaches to strengthen the usability and security of password authentication systems using natural language processing.

Thomson Reuters Labs is the applied-research arm of Thomson Reuters, working with some of the world's most comprehensive legal, tax, and corporate datasets to advance AI for professional services.

Over an 8-month internship, I contributed to a legal text entailment research project and a named-entity recognition product initiative:

• Co-authored two peer-reviewed papers exploring prompt engineering techniques for large language models on legal reasoning tasks.
• Collaborated with research scientists to identify opportunities for applying state-of-the-art NLP methods to legal products.
• Evaluated zero-shot, few-shot, chain-of-thought prompting, and fine-tuning strategies across GPT-3 and T5 using the Hugging Face and OpenAI APIs for domain-specific reasoning.
• Benchmarked baseline and state-of-the-art models — including spaCy, Conditional Random Fields, and LegalBERT — on a highly imbalanced named-entity recognition dataset.
• Maintained rigorous documentation of literature reviews, data processing, and experimental results following lab standards.

The AI Hub at Durham College partners with industry to deliver AI solutions that uncover business insights and drive productivity and growth.

• Built interactive dashboards on historical gold-market data using Tableau.
• Implemented machine learning models for time-series forecasting, with end-to-end documentation covering datasets, algorithms, APIs, data-flow diagrams, and optimization options.
• Presented data-driven insights and recommendations to business stakeholders.

Ontario Tech University (UOIT)

GPA: 4.24 on a scale of 4.30

Location: Ontario, Canada

Durham College

GPA: 4.83 on a scale of 5.00

Location: Ontario, Canada

Donghua University

Location: Shanghai, China