- Trained and deployed a GPT-based model to generate enhanced titles from original ones across multiple categories. Achieved 84% win rate and 13% tie rate in a side-by-side comparison. Collected and verified dataset results through crowdsourcing and conducted various experiments with prompts, model parameters, and datasets for improved performance and accuracy.- Migrated from local Alibaba data system (ODPS) to Hadoop.- Developed a KNN pipeline for mapping unorganized sellers' attributes to canonical ones. Used NER embeddings of the attributes and utilized FAISS to identify the nearest canonical attribute for each unorganized attribute.- Read papers by different themes and presented them on Reading Club multiple times.- Designed and deployed a solution for ranking user reviews on local Russian items with the objective of promoting positive reviews. Implemented a heuristic approach that considered review rating, predicted rating, length, number of images, etc.