Hi, I'm Sohee!

Shaachi, K2 Corp.

As a software engineering intern at Shaachi, I contributed to developing and integrating multiple features of their product. As my first time working at a startup, I gained valuable experience and insights into product development and dev ops. From enhancing machine learning algorithms for targeting features to AWS(S3, API Gateway, SQS, EC2) and API testing, I developed and tested backend platforms for Shaachi.

• Learn more

WordLab

Using pre-trained BERT models, we created an open-source research software for calculating surprisals for multilingual A-maze trials. Initially, the goal was to develop a more accurate model for Mandarin maze trials as Mandarin's orthography requires different tokenization. Upon completion, we expanded the code to use other HuggingFace models, such as multilingual BERT and GPT-2.

• What are surprisals?
• Code to Multilingual A-maze

Image Classifier (ML Project)

Using PyTorch, I built a CNN to classify dog breeds from images. Specifically, I focusd on distinguishing Golden Retrievers from Collies. In the process, I experimented with transfer learning and data augmentation to enhance the accuracy of the algorithm.