Corleone Yang

Baynovation: As a Software Engineer, I developed ML workflows on AWS SageMaker, using clustering for feature selection and Transformer-based models for loan default prediction. I integrated these into RESTful APIs with Flask and deployed them to AWS EKS, improving prediction accuracy by 18%. Additionally, I developed an AI-powered job application tool using Django and RLHF (Reinforcement Learning from Human Feedback), deployed on AWS EC2. I also built a Django-based investment simulation web app for ETF return analysis, integrating historical data processing and predictive modeling to simulate returns for popular ETFs.

Tencent: As a Machine Learning Engineer, I constructed a machine learning environment using Conda and Ubuntu 20.04 LTS. I developed a reinforcement learning framework with OpenAI Gym and PyTorch to implement the Deep Q-Network (DQN) algorithm, achieving an 86% task completion accuracy for smart agents in simulated environments. Additionally, I designed a Transformer-based recommendation system using production test data, implementing a sequence-based model using PyTorch to process user click data, resulting in a 7% improvement in click-through rate (CTR) in the test environment. I also optimized model training pipelines using CUDA for GPU acceleration, reducing training time by 13%.

XiuNeng Capital: As a Quant Research Intern, I designed and implemented an Event Study Algorithm using Python, pandas, and matplotlib to analyze the impact of specific events on stock prices. This improved event impact prediction accuracy by 18% through integrating data from Bloomberg Terminal and internal databases. I applied statistical methods and machine learning techniques for factor mining, including correlation analysis, PCA, and cross-sectional regression. I incorporated significant factors into a multi-factor model for portfolio construction and back-tested the strategy using Python, achieving a 12% increase in risk-adjusted returns compared to the benchmark.