Hello, I am Thoranna

Worked with a group of researchers to create WineSensed, a large multimodal wine dataset with 897k images and 824k reviews covering 350k unique bottlings. Developed a concept embedding algorithm that combines human experience with machine learning to align with intricate human perception of flavor. Published at NeurIPS 2023 (paper) and featured in the Telegraph (article).

Worked as the lead AI engineer on a team that delivered the beta of an AI coding platform for kids, gaining 400 users in 3 months. Kids learn to create apps and games using natural language. Built a multi-model LLM solution with code execution, structured output, and conversational AI. Video shows a game generated on the platform.

I worked as a part of a team of students for Genki Instruments developing an algorithm that could read data from simple motion sensors and build a classifier that recognizes the movements. During this project I learned valuable skills in machine learning fx. tuning a neural network, pre-processing data for different classifiers and multiple evaluation methods for classifiers. Video shows me trying out the classifier in real time on the hardware ring by using gestures to control bluetooth devices in the office.