Hey! I’m Sinem, a 19-year-old based in San Francisco, California.
I’m doing a double major in Physics and Computer Science (Data and Statistics) at Minerva University.
For over a decade, I have focused on building robots and working on programming, materials science and space technologies.
This portfolio contains selected technical projects I have built or contributed to.
Let’s connect!
Socials: LinkedIn | Github | Medium
Email: sinem.unlu@uni.minerva.edu
An Augmented Reality tool that helps people with dyslexia read up to 60% more fluently by adapting a written text’s features like font, size, spacing, color, and background color. Built using Brilliant Lab’s Frame glasses, the system captures physical text via the integrated camera and uses Python-based OCR to process the content. The reformatted text is then displayed on the Frame’s OLED display to provide a customized reading experience. I developed the end to end technical pipeline and a web interface that allows users to personalize visual settings to their specific needs. This solution bridges the gap between digital accessibility and the physical world by offering a portable alternative to traditional reading aids.
I am working as a student consultant in a year-long partnership with Google to develop an AI accountability coach designed to increase goal attainment. Using Google Gemini’s 2.5 Flash model as a base, I built a system that utilizes voice interaction and proactive questioning to help users balance mental well-being while tracking their objectives. The coach features a fully adaptive personality where traits, voice, and response length adjust dynamically based on user input. In Spring 2026, my team and I will transition the project into a physical companion device designed for integration into personal living spaces.
As lunar exploration shifts toward long-term settlement via the Artemis Program, accurate regolith simulants are essential for testing In-Situ Resource Utilization (ISRU) and 3D printing. I developed AYRAN, a high-fidelity simulant that replicates the chemical and mineralogical properties of the Apollo 17 sample 72501 with over 98% accuracy. Using a Python optimization algorithm I designed, I was able to simultaneously match complex chemical and physical profiles, overcoming the limitations of existing bulk simulants. The material was characterized using XRD, XRF, SEM, and TGA at Sabancı University, providing a more reliable foundation for terrestrial research into lunar habitat construction.
**
For FIRST Robotics Competition’s 2024-2025 Season, I was the team captain for teams #9232 and #10328. I designed the CAD of our robots and worked on building the mechanical and electrical components. I primarily used Onshape and Fusion 360 for CAD, integrating multiple subsystem designs, including intakes, lifts/elevators, chassis, and outtakes.
For the RoboChallenge Humanoid Sumo Category, I’ve built a small humanoid robot and worked on automation of the system. My team won 3rd place internationally in 2024!
I also designed and programmed 2 firefighter robots using LEGO and Arduino that won Global 2nd and 4th Place in Robotex International 2023!
https://drive.google.com/file/d/1aBR2q7JFQzPlnIylU8Nblpko1TSSPE3-/view?usp=drive_web
As of 2026, 68% of web pages have broken HTML structures, causing 71% of users with disabilities to abandon inaccessible sites immediately because they cannot effectively use voice assistants.
Voxly simulates how screen readers and voice assistants navigate a website by running six core accessibility checks (heading hierarchy, navigation labels, form/button labels, link context, and skip links) then providing AI-suggested code fixes and live audible demos.
How we built it: The React (Vite) frontend manages user flow and displays test progress. A Node.js backend uses Daytona to run headless browser DOM analysis, extracts headings, links, and forms, passes issues to CodeRabbit for AI fix suggestions, and converts explanations into audio using ElevenLabs.io. Sentry logs errors during analysis and demos. Finally, we also generate a breakdown of the issues using Daytona.
