Ruotong Wu

About Ruotong Wu

Highly organized and passionate software professional with 4 years of experience in…

Experience

1. Software Engineer

   Microsoft

   Aug 2021 — Present

   Designed and implemented MicroAnswer triggering service from scratch to decide optimum answers provided on bing.com, resulting in 5.34M DAU (C#)\\\\n- Designed and implemented an automated training pipeline, which helped partners save at least 50% of iteration time during training, testing and validating models, already have 2 partners onboarded within one month of launch (Azure Data Factory, Azure Function, Azure Data Lake)\\\\n- Designed and implemented a feedback loop to improve the quality of answers on Bing by statistically analyzing user feedback, resulting in a 0.11% increase in daily active users for recipe and job-related answers (SQL)\\\\n- Maintained the stability of the internal rule-based NLP platform Dolphin which supports over 100 partners to correctly trigger their answer in bing.com\\\\n- Created pipelines to train and evaluate models using XLM-RoBERTa and Microsoft DeBERTa models, improving the Bing Q&A experience and added 4.5% in terms of precision, and maintained the stability, portability and robustness of online serving\\\\n- Designed and implemented end-to-end pipelines, and web apps to interact with managed/crowdsourced judges to generate multi-language Q&A data. Resulting in a daily increase of data points with a precision rate of 91% or higher (WebApp, HTML, Javascript)\\\\n- Improved the Content Publish Date signal in Document Understanding pipeline by 32% in terms of precision, and generated new signals: Webpage Visual Date and Visual Time, to aid downstream processing (C#)\\\\n- Designed and implemented internal tools to visualize the models trained by the team for internal colleagues\\\\n- Worked with team to migrate the internal streaming pipeline from self-built platform to Kafka, largly improved the E2E latency\\\\n \\\\n\\\\n \\\\n \\\\n\\\\n \\\\n