I’m a 3rd year Computer Science student at Toronto Metropolitan University
with a minor in Mathematics. I currently work as a Data Scientist at
TMU and as a Senior Research Assistant in TMU’s CMIC Lab. My research
includes UM-GCN, a graph network for chest X-ray classification
submitted to SPIE Medical Imaging 2027. Outside the lab I build tools
like Pagenova, a Chrome AI assistant, and an F1 race-prediction
pipeline with XGBoost.
Formulated standardized data annotation protocols and
automated pipelines for multi-planar ultrasound sequences,
building high-quality ground-truth datasets to train and
benchmark multi-view segmentation models.
Engineered spatio-temporal deep learning segmentation
models to track dynamic needle trajectories across 3
orthogonal ultrasound planes over time, advancing
real-time precision for prostate cancer interventions.
Engineered a Python automation pipeline that translates
and validates data inputs for reactor simulation software,
reducing manual data preparation time by 70% and
eliminating transcription errors across critical safety
workflows.
Processed and mapped 10,000+ structured records per run,
enabling consistent, audit-ready data handoffs in a
classified, federally regulated nuclear safety
environment.
— to Present
CMIC Lab, TMUSenior Research Assistant (Feb 2026–Present)
Research Assistant (Oct 2025–Feb 2026) · Toronto,
ON
Collaborating with a multidisciplinary research team to
study the social dynamics of computer-mediated
communication (CMC).
Analyzing behavioural and communication data using
statistical tools to identify patterns related to
deception detection and social prediction.
Originated a 40% increase in sales by using creative
social media marketing techniques, which helped the
business become more visible in the marketplace.
Maintained and strengthened ties with more than 13 B2B
clients throughout Canada, resulting in steady revenue
growth, good cash flow, and the creation of a distinct
product demand based on local geography.
Worked on developing persuasive sales pitches according
to the changing season by addressing customer questions
and objections, resulting in tailored solutions to
customer needs, which led to consistently meeting weekly
sales targets.
Rewarded with the best salesperson award for June,
generating over 120% of the weekly sales goal.
Performed in-depth market research, which enabled the
company to expand into the USA with success, gaining a
sizeable market share, and strategically positioned the
business.
Worked in tandem with shipping and operational partners
to optimize packaging, improving customer happiness and
cutting expenses by 6.2%, all of which helped to
streamline logistics.
— to
Boron DiamondsFounder · Regina, SK
Dived into my start-up of an online drop shipping store
of imitation jewelry serving over 400 orders with an
average order value of $23 with healthy reviews.
Worked closely with affiliate marketers and social media
influencers to drive sales, and increased sales by
targeting potential customers through analytical
performance marketing.
— to
Bombay SpicesAssistant Store Manager · Regina, SK
Implemented strategic marketing techniques that generated
a $2 million increase in net sales, while managing
warehouse operations, inventory, and cross-training new
associates.
Improved team efficiency by 15% through optimized
stocking strategies and maintained strict safety
protocols while operating heavy equipment like forklifts
and cherry pickers.
Designed tailored lesson plans, custom practice
assessments, and targeted reading comprehension
strategies to improve student knowledge, boost
self-confidence, and reduce test anxiety.
Evaluated individual learning needs and academic progress
while supporting classroom instruction through one-on-one
tutoring, homework preparation, and regular performance
feedback.
Projects
PagenovaChrome AI assistant Extension · React, TypeScript,
Vite
Built a Chrome extension that grounds streaming AI chat in
open tabs, uploaded documents, and images, supporting local
Ollama and cloud providers (OpenAI, Anthropic, Gemini,
OpenAI-compatible).
Designed a budget-aware RAG pipeline (chunk → embed →
retrieve) with on-device Transformers.js embeddings,
IndexedDB vector storage, and automatic full-context vs.
retrieval mode based on model window size.
Implemented a React UI over a typed service-worker protocol
with abortable streaming, session-only API keys, SSRF-safe
image fetching, and a 120+ test Vitest suite covering
providers, retrieval, and streaming edge cases.
Preprocessed and analyzed free practice session data from
the FastF1 API (driver speed, lap times, conditions) to
build a structured dataset for predictive modelling.
Trained an XGBoost model to forecast driver speeds and race
performance for current-season events using last year’s race
data as benchmarks, improving prediction accuracy through
feature engineering.
Designed UM-GCN, a multiscale label-attention graph network
that builds image-specific disease graphs from attention
overlap, replacing fixed co-occurrence statistics with
entropy-modulated message passing.
Achieved macro-AUROC of 0.8318 and macro-AP of 0.3226 on
ChestX-ray14, outperforming a tuned EfficientNet-B4 baseline
through controlled ablations, with statistically significant
AUROC gains on 5 disease classes (Atelectasis, Infiltration,
Pneumonia, Edema, Emphysema) confirmed via external
validation on CheXpert.
