I'm Xinqi Guo.

I lead advertising strategy development for our Amazon store by using data-driven methods to increase performance and efficiency. Within three months, I improved ROAS by 240% through continuous campaign monitoring and optimization. I designed a keyword optimization framework using brand and competitor insights to maintain a conversion rate above 10%. I also integrated Amazon order data, customer behavior, and market basket analysis to personalize campaigns and quadruple new customer acquisition. Additionally, I engineered a real-time, serverless bidding and monitoring system using AWS services such as Glue, Lambda, S3, and SNS. This system dynamically adjusted bids based on hourly ROAS and conversion trends and automated campaign tracking with AWS Glue Crawlers and ETL processes.

During my internship at InnerView Group, I developed an alignment score metric to quantify perception gaps between frontline employees and leadership on competitive positioning. I used Gaussian Mixture Modeling (GMM) to segment survey respondents based on tenure, demographics, and responses, revealing patterns in advocacy and perception. I also built interactive Tableau dashboards that visualized advocacy trends across roles and regions, providing strategic insights to guide internal communications and training initiatives.

At IRIS, I supported the quality and integrity of the 2024 UMETRICS Dataset, enhancing six major datasets that covered over 580,000 sponsored research grants from 100+ universities. I developed an algorithm that improved data linkage using customized stopwords, groupwords, and abbreviations, minimizing dependence on text similarity alone. My approach reduced manual review time by 50% by creating differentiated cleaning procedures for organizational and individual data.

At Furtrieve, I enhanced the performance of an animal activity detection algorithm by reducing behavior detection time to under five seconds and improving the model’s F1 score by 10%. I achieved this through optimized training and data labeling using DeepLabCut with a ResNet backbone. Additionally, I streamlined customer response processes by summarizing customer feedback with OpenAI APIs and managing customer satisfaction surveys.

During my time at IBM, I worked on image classification using Python libraries like PIL and OpenCV to automate preprocessing steps such as reading, cropping, rotation, and scaling. I created hybrid images using high-pass and low-pass filters and implemented HOG and LBP feature extractors. These features were used to train a model to distinguish between cats and dogs, combining computer vision with machine learning techniques.

At Tencent Games, I conducted user behavior analysis for Apex Legends by mining player feedback from Bilibili and Reddit, uncovering trends that led to a 4x increase in new player identification. I also discovered key market differences between Chinese and Western audiences and proposed budget reallocation strategies accordingly. In addition, I created a prototype model to predict daily PUBG user activity based on engagement, news, match schedules, and streaming patterns.