Beijing, China
Founding Engineer at Sheet0, building Helio. I work on AI-native products, agents, and workflow automation: turning messy, high-friction tasks into reliable software that helps people move faster. At Sheet0, I’m focused on building Helio from 0 to 1 across product, engineering, and infrastructure. That includes designing agentic workflows, building full-stack systems, improving user experience, and shaping the foundations that make the product useful, fast, and dependable. I’m especially interested in the intersection of AI agents, product taste, automation, and real user feedback. My favorite work is taking an ambiguous problem, finding the sharpest version of it, and shipping something people actually want to use.
Building the future of AI workflows at Helio: https://www.helio.im/
Rotation 1: Data Analytic Transformation Team, Car Service Business • Project Focus: Developed an end-to-end solution to detect and mitigate abnormal activities in Volvo’s VIDA (Vehicle Information and Diagnostics for Aftersales) system. • Key Contributions: • Designed and implemented a full-stack automated dashboard and deployed on Azure Web App Service, enabling streamlined monitoring, blocking, and whitelisting of VIDA users. • Enhanced regulatory compliance by safeguarding data privacy and adhering to local governance standards. • Secured business revenue by preventing unlicensed system usage and blocking unauthorized access attempts. • Impact: Reduced regulatory risks, improved system security, ensured compliance with data governance policies, and protected revenue streams by mitigating unauthorized system usage.
Earnings Call Q&A Analysis: Utilized BERTopic topic modeling to deeply analyze the Q&A sessions of earnings calls, identifying common question patterns through HDBSCAN hierarchical clustering techniques. Also, employed the c-TF-IDF method for efficient keyword extraction, revealing the core themes and trends of the discussions, providing richer insights for investment decisions. Financial Market Analysis QA Chatbot: Developed a QA Chatbot using RAG and LangChain, aimed at providing instant financial market analysis and insights to users by analyzing U.S. Securities and Exchange Commission (SEC) filings, earnings call transcripts, and the latest financial news.
• Developed an NLTK, Word2Vec, and MLP-based text classification pipeline to predict client industries from employer names, achieving F1-scores of 0.72 (level 1) and 0.8 (level 2). • Created a PySpark script to clean and update client industrial information, increasing data coverage from 65% to 90% in collaboration with product and risk management teams. • Clustered 8,000 survey data points using PCA and KMeans algorithms, identifying three main customer loss groups and providing visualizations for the product team to pinpoint primary reasons.