student · developer · problem solver — 43.46°N 80.52°W
Excited to meet you.
A student and developer headed to Software Engineering at the University of Waterloo — building agentic AI systems, writing plain-English walkthroughs of ML papers, and competing in math and programming contests.

~/experience
ML Forward Deployed Engineer (Intern) · Cimba.ai
May 2025 - Sept 2025 · Remote/Seattle
Embedded with a Fortune 500 rental company, scoping and shipping agentic workflows into their production stack. Integrated AWS Bedrock with Snowflake so agents could act on enterprise data, and designed autonomous playbooks with Bedrock LLM tool-calling that lowered cost by 18%.
Full-Stack Developer (Intern) · MotionShark
May 2025 - Sept 2025 · Remote/San Francisco
Shipped 40+ UI enhancements to the collaborative whiteboard, prioritized by user interaction metrics. Moved latency-critical backend logic to AWS Lambda, cutting 0.3 seconds off the core whiteboard feature, and set up GitHub Actions CI/CD with tests on every PR and auto-deploys.
Founding Engineer · FinTech Nexus
Aug 2024 - July 2026 · Remote
Led 5 engineers to architect and launch fintechnexus.ca on React, Next.js, and Flask, owning design and deployment. Launched The Nexus Foundry, a 1,000+ student incubator integrating Ultra (YC25) and Vogent (YC25).
Frontend Developer (Intern) · Zeuty
Jun 2024 - Sep 2024 · Waterloo, ON · Remote
Cut app load time by 30% by reducing initialization work and trimming bundle overhead. Developed cross-platform modules in React Native, Expo, and TypeScript on native device APIs for iOS and Android, and implemented AWS-backed authentication with restricted-user access controls.
~/education
University of Waterloo, Software Engineering (BSE), Co-op
2026 - 2031 · Waterloo, ON
Ontario Professional Engineers Foundation Entrance Scholarship ($2,500, 1 of 2 in Software Engineering). President's Scholarship of Distinction ($2,500). Engineering International/Research Award ($1,500).
St. Marcellinus (Grade 12)
2021 - 2026 · Mississauga, ON
Top 6 Average: 98.25%
~/projects
Pre-registered eval harness measuring per-capability damage in GGUF-quantized LLMs across 7 bit-widths. Found that the damage is precision-dependent: at 3-bit, retrieval takes about 3.2× the damage arithmetic does. Paper in prep for ICLR.
Neural audio retrieval engine — a semantic Ctrl+F for sound. Maps raw waveforms into a 512-dimensional embedding space with LAION-CLAP so you can search unstructured audio for sonic signatures (glass shattering, sirens, a spoken phrase) instead of just transcripts.
AI-powered post-operative recovery assistant that helps patients find nearby pharmacies and manage their recovery after surgery.
Security application that detects car theft and automatically sends alerts to police and car owners with customizable daily protection features.
Innovative messaging platform using unsupervised machine learning to detect and prevent child trafficking attempts, targeting Southern African countries.
Agricultural solution that analyzes plant deficiencies to revitalize soil quality, helping communities diagnose crop issues efficiently during climate challenges.
~/honors
Canadian Computing Competition (CCC)
Senior Top 5% Distinction (Score: 43)
February 2025
CMIMC (Carnegie Mellon Informatics and Mathematics Competition)
Rank 23 internationally
April 2025
Jane Street Estimation Competition
3x Winner
April 2025
AIME Qualification
AMC 12B (2024, 2025 Top 10% globally).
February 2025
Grade 12 Euclid Mathematics Contest
65 (Highest in School + Written one year)
April 2024
Grade 11 Fermat Mathematics Contest
128 (Top 7% internationally)
February 2024
AP Calculus BC
Score: 5
May 2025
AP Scholar with Distinction
Completed 9 AP Exams (5,5,5,5,4,4,4,3,3)
May 2024
~/writing
August 8, 2026 · 40 min read · PDF
4-bit quantization really is nearly free — but at 3 bits both long-context retrieval and multi-step arithmetic break, and retrieval takes 3.2× the standardized damage. Eight components get better when you quantize them. And the standard way of ranking which layers matter turns out to fail four separate statistical tests. A plain-English walkthrough of the paper, with every figure and number intact.
August 2, 2026 · 22 min read
How a thirty-gigabyte AI model gets rounded down to fit on a laptop — scale and zero-point, calibration, granularity, integer-only inference, QAT, and the outlier problem that forced large language models to invent new tricks.
~/volunteering
Program Lead + Head Secretary
For The Love of Reading (Brampton) · Sep 2022 - Present · Mississauga, ON
Teacher
Taught my own grade 4 class. · September 2022 - January 2023 · Remote
Speaker
TEDxMcFarren Boulevard Youth · July 2025 · Remote
Speaker
HealthCareCoding · September 2022 · Remote
~/certifications
Open to internships, research, and good conversations.
