Katerina Dimitrova
Hi, I'm Katerina — you can call me Kat, like KitKat 🍫
Welcome to my little corner of the web,
I'm a junior student-athlete majoring in Computer Science and minoring in Engineering Leadership Development at Penn State. I'm also a member of the women's tennis team. I'm passionate about building and exploring different ways to solve problems. I'm fascinated by how technology and machine learning can be integrated into areas like healthcare to create meaningful impact and improve people's lives.
Over the last few months, I have been developing projects such as a real-time ASL-to-Text translation system and an AI Cognitive Distortion Detection system. These projects have given me hands-on experience in computer vision and NLP, taking me through the full pipeline — from data collection and feature extraction to model development in PyTorch and the use of LLMs. Additionally, building a book-selling website using Figma and Next.js strengthened my skills in UI/UX design and web development. I'm driven by how every project brings new challenges and opportunities to grow.
Projects
(2nd Place, Nittany AI Challenge)jan 2026 - present
- Developed a real-time ASL gesture-to-text translation system using OpenCV for frame capture, MediaPipe for landmark extraction, and a Temporal Vision Transformer for sign recognition, streaming live captions via virtual camera integration
- Increased model balanced accuracy from 51.5% to 89.7%, with further model refinement improving it to 94.8%
- Developed a vision screening application using Python, Streamlit, and OpenCV, implementing an adaptive binary search algorithm to estimate refractive error with ±0.25 D precision
- Built a real-time computer vision pipeline using OpenCV, NumPy, and Pillow to detect a credit card via HSV color masking and estimate user-device distance using a pinhole camera model for accurate calibration
- Using a locally served LLM (via vLLM) to analyze diary entries from post-concussion patients to detect and classify cognitive distortions in their self-reported thoughts — determining whether a distortion is present and identifying its type
- Investigating how cognitive distortions in patient-generated text relate to concussion symptom burden and recovery outcomes, aiming to identify thought patterns that may perpetuate or predict persistent post-concussive symptoms
- Conducted user research with the client to understand goals, informing website design and functionality
- Designed UI/UX in Figma; built the website using React, TypeScript, and Node.js
- Integrated third-party APIs for payment processing (Stripe, Przelewy24), automated email delivery (Resend), and database management (Supabase)