I'm an EECS student at UC Berkeley with a deep interest in artificial intelligence and machine learning. I'm currently working at the Berkeley Artificial Intelligence Research (BAIR) lab, where I'm exploring the security implications of Large Language Models.
I'm always eager to learn and apply my skills to build impactful solutions.
Can an AI find hacks that humans miss? We are currently exploring how large language models behave when you drop them into the world of Android security — a space full of 'hidden' vulnerabilities. To really push their limits, we build controlled environments where models can experiment and sometimes succeed in surprising ways. It's a glance into what AI-powered security might look like in the future.
At the Hybrid Ecologies Lab, I worked on a project that combined physics and art. We explored Moiré patterns — those shifting lines you see when two grids overlap — as a way to build ultra-sensitive capacitive sensors. I helped prototype a tunable sensor and refine the signal processing so that tiny environmental changes became measurable. I also built visualization tools that let researchers see sensor behavior in real time, transforming raw capacitance data into something you could immediately interact with and understand.
I built a Chrome extension that brings course and instructor reviews directly into the academic portal, so students can see real-time feedback without leaving the page. Using Chrome Extension APIs, it runs securely and smoothly. You can check it out on the Chrome Web Store.
Back home in Myanmar, many shops run on POS systems but don't really have tools to understand their own sales data. This summer, I was fortunate enough to access a store's database. So I dug into it, reverse-engineered the schema, and started building a dashboard that could actually highlight trends in pricing, inventory, and sales performance. I used Python, SQL, and a bit of machine learning to uncover patterns that the POS couldn't show. The write-up is in Burmese, and I haven't shared the full analysis online (since the data is private), but if you're curious I'd be happy to walk you through some of it.
This project was basically to get myself into classes that get filled up quickly. I built a system that scrapes course schedules, faculty info, and grade distributions from the college portal, then organizes it all in a structured SQLite database. By containerizing the scraper with Docker and automating interactions using Selenium, I could efficiently handle thousands of entries in real time. The setup also feeds data to a Discord bot, where I send queries to retrieve course's enrollment statistics.
At Cookie Academy in Yangon, I built a tool that makes content creation faster and smarter. I built a Python tool that turns structured data into dynamic HTML and CSS templates, streamlining the workflow and boosting efficiency by 40%. Along the way, I wrote clear documentation so the team could easily maintain and extend the system.
I have some thoughts and notes (drafts) on various topics. Check out my notes.