New York, New York, United States
• Applied deep reinforcement learning to alpha research, focusing on developing strategies that optimize factor discovery • Trained an agent using a multi-layer perceptron and the PPO algorithm to generate alpha factors. Delivered a feature extractor using LSTM and Transformer on daily frequency stock market data • Trained LASSO, XGBoost, and LightGBM to fit the generated alpha pool of size 200 and monitor their performance on daily, weekly, and monthly forecasting tasks using Tensorboard • Fine-tuned 65 selected factors and tested the alpha pool using qlib. The best pool produced a 15% IC mean and 3.3 ICIR on the test set. The backtest result benched a 1.8 Sharpe ratio, 48% annual excess return, and 10% max drawdown from January 2024 to August 2024 on the CSI500 data (out-of-sample)
• Delivered scalable full-stack services, including a custom PDF export module and remote storage service for client companies’ timesheet management. Delivered user interfaces enabling 30+ companies to track employee status in real-time • Designed and implemented restful API endpoints using Kotlin and PostgreSQL, developed cache mechanism for existing API and reduced average response time by ~500ms • Used Flutter to develop a cross-platform UI and Riverpod for reactive caching and state management.
• Designed and built a full-stack large language model application that automates the publication, editing, and management of equity research reports • Developed an AI search tool for company internal data using Langchain as frontend and ChatGLM as the backend LLM model • Extracted metadata from equity research reports into Elasticsearch to enrich search results • Developed a multi-platform UI using Flutter, ensuring a seamless cross-platform experience on both mobile and desktop devices