Focused on speech and natural language processing products, participating new product build and development 1. Boost the performance automatic speech recognition service, addressing issues of hallucination, text repetition, unexpected translation, among others. 2. Designed LLM prompting framework for long-text analysis and application of customized scenarios. 3. Developed methods/metrics to evaluate the credibility of natural language generation. 4. Design load balancing pipeline to reduce the pressure of limited GPU resources. #Whisper #OpenAI #Azure #GCP #Elasticsearch #Fastapi #Docker #Kafka #Redis

Works Focused on research & development of machine learning products on e-commerce service and business 1. Built the end-to-end machine learning pipeline for data scientists to facilitate ML product development with MLOps, reducing >50% PoC time. 2. Developed generative language models on automatic copywriting application for the push service and social media marketing. 3. Trained computer vision models on fashion images to extract information from product items, which helped making tags on users and items to boost the e-commerce service. 4. Designed recommendation algorithms based on the user logs and product tags a. raise the click-through rate for new users with ~119%. b. gain the opening rate of push service with >30%. #GPT2 #HuggingFace #AWS #Sagemaker #Lambda #Chalice #PyTorch #PySpark Competitions 1. 7th place (top-1%) in Shopee Code League 2020 for e-commerce product categorization. 2. Bronze medal in Kaggle competition. https://www.kaggle.com/radream Projects/Codes can be found on https://github.com/yurayli