Hi there,
I'm Saı⚽.

Capital One

Toronto, ON

Engineering an end-to-end interview-scheduling system using Next.js, Go, PostgreSQL, and AWS, bridging candidate and recruiter workflows through dynamic availability mapping, SSO integration, and centralized job-round management.

University of Cambridge · University of Waterloo

Cambridge, England

Engineered active-learning pipelines for GEDI biomass estimation, demonstrating that spatial sampling mitigates the cold-start problem by outperforming random baseline strategies by 27% under extreme data starvation (N=25). Built a 2D landscape generator using Gaussian Random Fields to simulate 3,000+ data points, modelling both smooth and chaotic environments to evaluate the situational superiority of different learning strategies. Architected a Python/SciPy evaluation suite to benchmark ML strategies, showing that Attentive Neural Processes (ANPs) outperformed Random Forests by more than 3× accuracy (R²) in chaotic environments under data starvation (<1% sample size).

Marsh McLennan

Toronto, ON

Automated case-handling by building an email classifier with RAG-based semantic search and a RESTful Node.js API. Resolved 1,000+ exposed secrets by engineering a GitGuardian/JWT remediation pipeline that won an org-wide tournament. Saved 100+ hr/wk by automating 300+-instance database deployments in GitHub Actions with rollback and secret handling. Bolstered preemptive security and development velocity by piloting a GitHub bot for OWASP vulnerability auditing and building a natural-language-to-workflow app via Mastra AI. Optimized reliability for 30+ apps by refining and documenting multi-cloud designs (OCI, AWS, Azure, GCP) alongside dev teams.

The Waterloo Alt Protein Project

Waterloo, ON

Extracted genomic insights for cultivated meat by architecting an end-to-end RNA-seq workflow to analyze public datasets. Optimized analysis of large-scale public single-cell datasets by engineering a reproducible Python pipeline (Scanpy, anndata, pandas, scikit-learn, rpy2) that enables efficient QC, clustering, and visualization of large datasets.

Mathematics Society, University of Waterloo

Waterloo, ON

Eliminated unreliable API dependencies for term lookups, stabilizing the interface via memoized utility functions. Streamlined admin workflows by engineering a multi-select UI and RESTful backend endpoints for bulk exam management.

Global X Investments Canada

Toronto, ON

Increased organic traffic to core financial pages by reducing load times by 15% and refining technical SEO — through API optimizations, front-end lazy loading, and metadata enhancements to maximize indexing and retention. Drove a 32.4% email open rate and scaled web presence by deploying 80+ pages and email blasts to support product launches. Saved 100+ hr/yr by automating EN/FR content synchronization and internal linking via custom Python and PHP scripts.

University of Waterloo

Waterloo, ON

Built and tested accessible web pages and tools within the LEARN LMS, streamlining QA processes and enhancing productivity through custom VBA solutions and improved internal tooling.

SlimeScholars

Remote

Engineered modular front-end features for a gamified learning platform using React.js and Tailwind, cutting latency through smart caching and preparing the product for alpha launch.