⚫ Maintained Forstack, a bespoke Signal‐as‐a‐Service product predicting movements in the financial time‐series using Temporal Fusion Transformation (TFT) model with an improved prediction accuracy (+1.3%) and wider financial data sources.⚫ Wrote articles on website and social media, covering various topics including computer vision, reinforcement learning, and methods for realizing greener AI to minimize carbon emission and explaining the technology details.⚫ Worked closely with treasury analytics team to make informed treasury decisions via data analysis and visualization.

Compared conventional game algorithms utilizing human knowledge of game rules as prior knowledge to deep reinforcement learning methods uisng a reward function and neural network models in the game Staircase Nim.Self‐studied Q‐Learning from scratch and implemented the algorithm using Python and Pytorch.Wrote collaboratively a paper published on PPRAI 2021.

⚫ Visualized a corpus of ancient fairy folk tales (1,000 tales in 7 languages) using the Stanford Core NLP packages.⚫ Examined the corpus from perspectives including sentence complexity, word frequency, and sentiment analysis with an independently cleaned and processed NLP datasets.

⚫ Proposed and implemented federated policy gradient and empirically proved the convergence of the model on Linear‐Quadratic Regulator (LQR) problem.⚫ Evaluated federated policy gradient’s convergence on more difficult cartpole problems under different parameter settings, and achieved better results compared with the vanilla one.⚫ Discovered that federated policy gradient could save up to 70% of the time and accelerated the whole process.

⚫ Participated in the initial part of an IPO program for an aluminum alloys castings manufacturing company, mainly responsible for the financial part of the program.⚫ Assisted in writing the company’s prospectus, building models to perform financial analysis for company’s business trend and profiting prediction.⚫ Arranged daily team meeting between the IPO team and the company executives.

⚫ Conducted research on the target companies, wrote the financial and business part of the due diligence report by analyzing the target companies’ and other comparable companies’ financial information.⚫ Organized 3 On-site due diligence process, gathering more detailed information about the target companies from interviews with their executives.