Let's Me Tell You

My Story

I’m a builder who believes clean code and thoughtful data go hand in hand. I love creating tools that solve real problems, whether it’s engineering backend systems or uncovering insights from data.

My journey into tech began when my family’s small business lost a significant amount of money due to a missed tax filing—everything had been tracked manually. I paused school and taught myself Python, Flask, and the Google Sheets API to automate the process. That experience showed me how impactful technology can be, and I’ve been excited ever since to build products that make people’s lives better.

What might surprise people is that I do my best coding outdoors, literally touching grass. I was born in a coastal mountain area, so I’ve always felt a deep connection to nature. When I’m stuck on a problem, I take my laptop to a park or sit under a tree, and suddenly, things start to click. Nature helps me slow down, think clearly, and write better code. It reminds me that creativity doesn’t only live in IDEs or server logs, but also in stillness and space. I’ve even convinced some friends to join me, and we’ve had a great time coding together outside.

Explore My

Experience

Panasonic North America

Spring & Summer Intern | February 2025 – July 2025

Implemented Python (Pandas) pipeline to parse unstructured JSON and PDF invoices into structured CSVs by applying regex-based field extractors and schema normalization, achieving 95% extraction accuracy
Transformed 18 raw tables with 5M+ rows using PySpark on Microsoft Fabric to build a centralized Lake Database of customer and tax data, enabling targeted loyalty strategies and reduced query latency by 25%
Built a serverless API on Azure using FastAPI to expose real-time KPIs computed from Lake Database, enabling Power BI to query live metrics and visualize customer behavior and tax transaction trends for key stakeholders
Built baseline classification models (Logistic Regression, Decision Tree, Random Forest) using scikit-learn to flag high-risk financial transactions, reducing false positives by 25% and contributing to an estimated $15,000 in cost savings.

Social NLP Lab, Drexel University

AI Research Assistant | Sep 2024 – Present

Applied NLP techniques including GPT-4 embeddings and zero-shot classification to analyze 10K+ Reddit memes, surfacing key shifts in sentiment polarity and slang usage with 85% topic coherence across identified discussion clusters.
Fine-tuned Large Language Models (LLaMA, GPT-3.5/4) for generating human-like explanations and sentiment summaries of meme content, enabling deeper sociolinguistic interpretation.
Designed and deployed a LLM batch inference system using OpenAI API and HuggingFace Transformers, enabling high-throughput processing of Reddit meme captions and reducing end-to-end labeling time by 35%.

KPMG

Software & Data Intern | May 2024 – Aug 2024

Containerized and deployed a high-performance C++ payment processing microservice to AWS EC2 using Docker, reducing deployment time by 60% and enhancing scalability and fault tolerance under concurrent load
Built CI/CD pipelines with GitHub Actions and AWS CodeDeploy for the internal finance dashboard app, enabling the client solutions team to deploy backend and UI updates seamlessly, cutting manual deployment time by 80%
Built Prometheus and Grafana dashboards to monitor request latency, error rates, and resource usage for payment portal APIs, improving system observability and accelerating incident response

Chase Cost Management

Engineering & Analytics Intern | Feb 2024 – May 2024

Refactored and redesigned pricing and message board pages using TypeScript, React, and Tailwind CSS, ensuring responsive design and improving feature adoption across 50+ client accounts
Increased testing coverage by 15% and backend reliability by 20% for Transaction Support Chatbot service by implementing more than 35 unit tests and end to end tests in JUnit 5 and Mockito

Turning Ideas Into Real Products

Discover My Portfolio

RefNet

Developed a platform that scrapes internships posted within 24 hours and automatically emails opportunities to 100+ students daily.

S&P500 Trading

Built a C++ console app to simulate long-term S&P 500 investments using historical Nasdaq data with buy, hold, and sell actions.

SummarAIze

Built a live transcription web app with React and Three.js, integrating Google Meet, Microsoft Teams, and Zoom for real-time meeting summaries.

betterTransit

Developed a machine learning model to predict passenger volume on buses and trains in New York at specific times using historical MTA data.

Facial Recognition

The project aims to develop AI facial recognition technology to enhance security for high-profile individuals such as singers from the general public.

ClickToConvert

This experiment aims to evaluate whether prompting students to specify their time commitment before beginning a free trial reduces dropout rates.

Janhi Ong