Yiyou LIN

Azure AutoML Platform - AutoML vision products such as image classification, object detection, video object tracking etc.; Azure OpenAI Finetuning team - LLMs customizations such as ChatGPTs (4o/4o-mini) knowledge injection (https://arxiv.org/pdf/2404.00213) & reasoning models (o1/ o3/ o4) reinforcement finetuning oai parity chasing.

Was owner of multiple projects. Developed deep learning models to tackle business challenges, and delivered solutions with highly scalable, highly available web services. • Smart Car Insurance （Computer Vision) • Text Content Censoring (NLP) • Comprehensive Image NSFW Detection (Computer Vision) • Content Understanding Systems for News Feed Recommendation (NLP & Recommendation Sys)

Worked on computer vision, CNN, deep learning related projects Familiar with: Keras, Tensorflow, OpenCV, Scikit-Image, Scikit-learn, spaCy • Scanned PDF Classifier, Project Owner - Identified document pages from scanned PDF documents - Used computer vision techniques to reduce OCR failures. • Property Image Classification, Project Owner - Build a transfer learning pipeline to identify room features. - Developed active learning framework to improve data quality.

Worked on the analytics team to model solutions for marketing problems. Accomplishments include: • Researched and developed a Python prototype to model dynamic marketing Return on Investment using Kalman Filter, which was an improvement from the traditional static RoI model using linear regression • Improved sparse regression model for TV rating prediction by 12% accuracy with Random Forest, Lasso Regression using R, which brough significant financial benefits • Designed testing schema and automated R codes to detect audience’s media preferences for high dimensional questionnaire with Chi-squared Test, Fisher’s Exact Test, Spearman Correlation Test, and Non-Parametric Bootstrap testing

Worked on customer segmentation and recommender systems. • Segmented customers by analyzing 2.9 million sales records and clustering 20 thousand customers’ behavioral data • Discovered three customer trends by analyzing 2.7 million transaction records and five years’ financial reports • Evaluated suitable recommender algorithms in Hilti’s context

Developed client communications, financial understanding and business sense. Provided audit service to 7 clients, and risk consultancy service to 3 clients in process of IPO. Auditing: • Detected skeptical financial accounts by cross-comparing data and analyzing business patterns for 6 projects • Appointed as lead positions due to excellent work. Enterprise Risk Service Consultancy: • Led Chinese team to conduct 3 IPO projects in the newly-set-up Enterprise Risk Service section • Reported over 20 risk issues by visualizing operation flows for purchase, human resource& financial department