Developed a pipeline for generating 1400+ synthetic pathological myopia fundus images and corresponding segmentation masks,
addressing the challenges of small and imbalanced image datasets.
• Utilized PGGAN, DGGAN, and diffusion models on Azure Machine Learning services to stream-line model development for the
creation of high-quality synthetic images, improving the availability of annotated data for eye disease classification.
• Improved dataset accessibility and balance by leveraging Kubernetes clusters for scalable model training, enabling more robust
automated tools in medical image analysis and eye disease classification.
• Tools used: Python, PyTorch Library

• Designed a Stream-lit based interface for the Autogen Library, enabling seamless AI agent interaction and proposal generation.
• Developed dynamic features to enhance and customize interactions between AI agents within an AI council, improving
collaboration and usability between the agents.
• Leveraged AWS Lambda functions to automate backend processes and enhance scalability and Firebase for real-time data
storage and authentication.
• Tools Used: Python, LLM, Stream-lit, AWS Lambda, and Firebase

Designed an app to locate and display available street parking using a real-time, interactive map, utilizing the Open Calgary
dataset to improve convenience for users.
• Implemented data processing and cleaning techniques to ensure accurate visualization of parking availability, completing the
project in 24 hours through effective teamwork and technical skills.
• Tools used: HTML, CSS, JavaScript, Python, Flask

Gained proficiency in machine and deep learning through collaborative research, creating a timeline and project planning with
team members and mentors.
• Developed an AI-powered study helper tool, integrating OpenAI API into a Streamlit-based GUI, parsing PDF data, and building
a vector database to enhance accuracy and relevance of responses.
• Tools Used: Python, Stream-lit, LLM, OpenAI API

Collaborated with a team of 20-25 peers to design and develop a Java-based vending machine GUI, simulating a grocery store
experience with seamless item scanning and payment functionality.
• Contributed to backend development by testing and debugging code, ensuring optimal performance, and demonstrating strong
teamwork and technical expertise to deliver a functional product.
• Tools used: Java, JavaSwing