Experience

Center for Intelligent Information RetrievalMay 2026 – Present

Research Assistant · Hybrid

Independent Study (Prof. Rahimi, 2026–27): Self-improving dense retrieval using hard-negative mining and LLM-generated training signals, evaluating iterative retraining on MS MARCO and BRIGHT with nDCG@10

AltereaJan 2026 – Aug 2026

Software Developer · Remote

Built teacher analytics for Agents of Influence, Alterea's AI media-literacy game for K-12 classrooms: rebuilt the React performance modal with collapsible skill breakdowns so teachers could read session insights at a glance
Restored correct behavior across classroom and account flows by resolving 10+ bugs in the frontend and Firebase backend, including form validation that wrongly required email consent and student-account display errors
Helped build the agentic pipeline behind multi-turn educational conversations

LeeYuen HousewaresJun 2025 – Aug 2025

Software Engineering Intern · Hong Kong

Replaced the purchasing team’s manual reorder estimates by building a demand forecasting model with scikit-learn on 10 years of sales history across 5,000+ SKUs for UK/US clients which was adopted by purchasing and sales
Cut OEM quote turnaround from 1–2 days to same-day by developing an internal React and Spring Boot quotation tool consolidating BOM costs, pricing across 55 suppliers, and production lead times into one interface
Replaced manual weekly competitor price lookups with automated daily refreshes by writing Dockerized Node.js scrapers on AWS tracking ∼ 1,000 SKUs across 6 sites, structuring output for procurement benchmarking

BioNLP LabDec 2024 – May 2025

Research Assistant · Hybrid

BioNLP Lab (2025): LoRA fine-tuned LLaMA, BioGPT, and Clinical-T5 on MedCalc-Bench to improve clinical-calculation accuracy, benchmarking against zero-shot baselines; presented at the URV Symposium

TruBridge HealthcareDec 2024 – Feb 2025

Data Science Intern · Remote

Scoped readmission modeling by analyzing 50K+ clinical records across hospital systems with pandas and SQLAlchemy, profiling relationships between readmissions and infection-control metrics to select predictors
Identified significant readmission predictors (p ≤ 0.01), including length of stay and prior infection incidents, by modeling readmission rates against infection-control KPIs with logistic regression and OLS in statsmodels
Cut clinical-record preprocessing from ∼ 20 minutes to under 5 by rewriting row-wise pandas operations as vectorized transforms in the AWS Lambda and S3 ETL pipeline, adding error logging for failed ingestion runs

ETL

Zyntra.ioJul 2024 – Aug 2024

Software Engineering Intern · Remote

Built real-time messaging with React, Node.js, MongoDB, and Socket.IO for 2K+ concurrent users
Integrated REST and GraphQL APIs, reducing render failures and response issues by 35%
Added JWT auth with token refresh and session expiry, cutting auth errors by 60%.