~Jack London
Co-led a project using open-source LLM "Qwen" to disable LLM generation on select topics while maintaining high performance. Performed rigorous experimental design and analysis. Leveraged GPT-4, MMLU benchmark, and BLEU/ROGUE scores as quantitative evaluation metrics (Python, PyTorch, Jupyter, Numpy, Transformers).
Developed a ML Ops Data Pipeline using AWS, Apache Airflow, and PySpark for a heart disease dataset. Conducted EDA, ETL, feature engineering, missing value imputation, followed by implementing and evaluating logistic regression and SVM models via PySpark and scikit-learn. Utilized Python, Boto3 for data loading, CloudWatch, S3 buckets, and RDS database.
Co-authored a paper on mechanistic interpretability (reverse-engineering LLM transformers), investigating emergent positional embedding of words in GPT-2. Performed data extraction, cleaning, and analysis; Conducted training with linear probing on a model with natural sentence data with Python (PyTorch, spaCy, Jupyter Notebook, TransformerLens), and created visualizations of results.
ML program for real-time license plate obfuscation. Implemented Computer Vision model YOLO, gaussian blurring, and evaluation metrics to optimize image utility while guaranteeing privacy of drivers. Used Python, PyTesseract, CV2, PyTorch.
Created and co-authored a research paper on a NLP model for cross-domain abstractive text summarization using incremental fine tuning on pre-trained models. Used Python, PyTorch, SBERT, HuggingFace Transformers.
Managed a team of four students, led the development of a health/medicine management app. Utilized an ensemble ML model to scan prescription bottles and built a GPT-powered assistant. Used React, Firebase API, Python, Google Vision API, spaCy, OpenAI ChatGPT API, Tailwind CSS, and Mantine.
I created a fully functional microinternship platform for Northwestern students, allowing students to apply to real micro-internship opportunities posted by startups and local businesses. Built with React, Tailwind, Firebase, Google Cloud, Typescript.
I built a program that detects brain tumors via an image segmentation model using Python ML Framework Keras, as part of a team project to create an assistive tool for radiologists.
During the Wildhacks Hackathon, I created a website that uses OpenAI GPT3, Multithreading APIs and Web Scraping to recommend gift ideas based on a short personal description.
Co-developed an interactive visualization of data on countries around the world (modifiable scatter plot, global map, bee-swarm plots) with D3.js, JavaScript, HTML, CSS. Iteratively designed with storyboarding for intuitive UI / UX.
I started this club as a freshman to gather students with or without coding experience to learn how to create mobile apps and build computational thinking skills. We learned to program in MIT App Inventor, Thunkable, Android Studio (Java), and Swift, and created the SMart School Supply Trading app and the WY Club App. We believe in App Inventor's vision that anyone can build apps that impact the world.
At Whitney Young, clubs are close-knit communities within communities. I've created a website to help extend the circle of connections, to allow students find infomation, passion and friends in over 130 clubs.
I gained lots of exposure to many areas of enginnering: I built a Stereolithographic 3D Printer, signed parts using CAD, programmed with Arduinos, soldered circuits to build radios, and visited Psyonic and the CAT Data Innovation Lab.
Northwestern Mechatronics Lab is a maker's playground. I had a lot of fun making my first autonomous vehicle: a smart shopping cart at CTD summer camp.
I started FIRST Lego League in 6th grade. Since then, I've been on a journey paticipating in FIRST Robotics Competition every year. FIRST Robotics has trained me to think like scientists and engineers.
Chicago is becoming a hacker's dreamland. I am a regular at ChiHackNight, a weekly event for Chicago coders to get together build, share and learn about civic tech. I worked on building a Chicago Virtual Assistant, a voice controlled 311 request system, and worked on the "Chicago Books to Women in Prison" Project. Here is a GitHub Repo to our project.
Doing overnight Hackathons is amazing. Despite being locked in old school buildings with hundreds of teenagers, the secret ingredients to achieve the magical wonder of these events are great ideas, caffine, code, more caffine, and a steady flow of chips and cookies.
5 Years Whitney Young Math Team
+ 5 Years Chicago Math Circle
+ 3 Years UCHI Young Schoolar Program
+ countless QED, NSML, CCML, AMC
∵ There are so many problmes waiting to be solved by us
A week of exploring Data Science through Nature, Art, Athletics, and Autonomous Vehicles. Learned to teach computers to distinguish data sets using many tools; JuliaBox, Jupyter Notebooks, Oculus VR Headset.
Summer 2017, I attended Fenghua Academy's Youth Leadership Program, spent 2 weeks living on campus with Chinese students, teaching English and organizing culture exchange activites. The Culture Exchange Summer Program for local high school students has provided great opportunities for international culture enrichment, volunteering, and leadership development.
Had the opportunity to meet industry legends, educators, new inventors, and nobel prize winners, who all gave inspiring speeches of incredible stories and impactful discoveries in the fields of science and technology, creativity and imagination.
I enjoy goofing off and engaging in recreation after a long day of school. Or, really, a long day of anything. I like making videos. Here are a few of the things I've made.