Experience

Harvard Kenneth C. Griffin Graduate School of Arts and Sciences

• United States
• Higher Education
• 100 - 200 Employee

• Grad Student

  • Sep 2021 - Present



University of Michigan

• United States
• Higher Education
• 700 & Above Employee

• Research Assistant

  • Oct 2019 - Jun 2020

  I was fortunate to work in the Strategic Reasoning Group supervised by Professor Michael Wellman, together with a group of most talented minds in the intersection of Economics and Computer Science. I worked on Strategy Exploration Problem of Empirical Game Theoretic Analysis. The main goal was to find Nash Equilibrium in large extensive form game where players' strategy is parenthesized by neural network. Specifically, my work includes: • Migrate EgtaOnline from Flux Cluster to… Show more I was fortunate to work in the Strategic Reasoning Group supervised by Professor Michael Wellman, together with a group of most talented minds in the intersection of Economics and Computer Science. I worked on Strategy Exploration Problem of Empirical Game Theoretic Analysis. The main goal was to find Nash Equilibrium in large extensive form game where players' strategy is parenthesized by neural network. Specifically, my work includes: • Migrate EgtaOnline from Flux Cluster to Greatlakes Cluster with Ruby Rails framework; • Disprove interleaving Double Oracle with Fictitious Play could lead to faster convergence in two player general-sum game; • Observe evaluation inconsistency in current literature when comparing strategy exploration algorithms, propose criterion to eliminate the bias and demonstrate its efficacy in synthetic and real world games; • Propose Minimum Regret Constrained Profile as an optimal evaluation metrics and perspective meta-solver and use Amoeba method to solve for the quantity; Show less I was fortunate to work in the Strategic Reasoning Group supervised by Professor Michael Wellman, together with a group of most talented minds in the intersection of Economics and Computer Science. I worked on Strategy Exploration Problem of Empirical Game Theoretic Analysis. The main goal was to find Nash Equilibrium in large extensive form game where players' strategy is parenthesized by neural network. Specifically, my work includes: • Migrate EgtaOnline from Flux Cluster to… Show more I was fortunate to work in the Strategic Reasoning Group supervised by Professor Michael Wellman, together with a group of most talented minds in the intersection of Economics and Computer Science. I worked on Strategy Exploration Problem of Empirical Game Theoretic Analysis. The main goal was to find Nash Equilibrium in large extensive form game where players' strategy is parenthesized by neural network. Specifically, my work includes: • Migrate EgtaOnline from Flux Cluster to Greatlakes Cluster with Ruby Rails framework; • Disprove interleaving Double Oracle with Fictitious Play could lead to faster convergence in two player general-sum game; • Observe evaluation inconsistency in current literature when comparing strategy exploration algorithms, propose criterion to eliminate the bias and demonstrate its efficacy in synthetic and real world games; • Propose Minimum Regret Constrained Profile as an optimal evaluation metrics and perspective meta-solver and use Amoeba method to solve for the quantity; Show less



UCLA Henry Samueli School of Engineering and Applied Science

• United States
• Higher Education
• 300 - 400 Employee

• Research Assistant

  • Jul 2019 - Sep 2019

  I worked in UCLA Machine Learing and Genomics Lab, supervised by Professor Sriram Sankararaman. Under the broad area of Transcriptome Wide Association Study (TWAS), my work focuses on mathematically extracting cell type specific gene expressions from the tissue-level data, and then performing TWAS on the extracted gene expressions features. Specifically, my work included: • Deconvolute tissue-level gene expression into cell-type specific ones with Tensor Component Analysis •… Show more I worked in UCLA Machine Learing and Genomics Lab, supervised by Professor Sriram Sankararaman. Under the broad area of Transcriptome Wide Association Study (TWAS), my work focuses on mathematically extracting cell type specific gene expressions from the tissue-level data, and then performing TWAS on the extracted gene expressions features. Specifically, my work included: • Deconvolute tissue-level gene expression into cell-type specific ones with Tensor Component Analysis • Perform TWAS on the cell-type specific gene expression and UKBiobank data • Optimize TWAS parameter estimation procedure to enforce sparsity on SNPs effect sizes • Develop data simulation scheme for gene heritability and correlation I especially enjoyed the vibrant research environment in Sriram Lab. The frequent cross-lab communications, cross-school communication all added to the creation of ideas. Show less I worked in UCLA Machine Learing and Genomics Lab, supervised by Professor Sriram Sankararaman. Under the broad area of Transcriptome Wide Association Study (TWAS), my work focuses on mathematically extracting cell type specific gene expressions from the tissue-level data, and then performing TWAS on the extracted gene expressions features. Specifically, my work included: • Deconvolute tissue-level gene expression into cell-type specific ones with Tensor Component Analysis •… Show more I worked in UCLA Machine Learing and Genomics Lab, supervised by Professor Sriram Sankararaman. Under the broad area of Transcriptome Wide Association Study (TWAS), my work focuses on mathematically extracting cell type specific gene expressions from the tissue-level data, and then performing TWAS on the extracted gene expressions features. Specifically, my work included: • Deconvolute tissue-level gene expression into cell-type specific ones with Tensor Component Analysis • Perform TWAS on the cell-type specific gene expression and UKBiobank data • Optimize TWAS parameter estimation procedure to enforce sparsity on SNPs effect sizes • Develop data simulation scheme for gene heritability and correlation I especially enjoyed the vibrant research environment in Sriram Lab. The frequent cross-lab communications, cross-school communication all added to the creation of ideas. Show less



SenseTime 商汤科技

• Hong Kong
• IT Services and IT Consulting
• 700 & Above Employee

• Research Intern

  • Feb 2019 - Jun 2019

  In the autonomous driving group, I focused in decision making model of the vehicle using Reinforcement Learning techniques. Given the waypoints along the road ahead, I help implemented two models: • End-to-end model. Trained Trust Region Policy Optimization and Generative Adversarial Imitation Learning driving control unit in Carla simulator with way-points as inputs and throttle-break and steer as output. Vehicle is capable of navigating in chartered town maps • Model-based RL… Show more In the autonomous driving group, I focused in decision making model of the vehicle using Reinforcement Learning techniques. Given the waypoints along the road ahead, I help implemented two models: • End-to-end model. Trained Trust Region Policy Optimization and Generative Adversarial Imitation Learning driving control unit in Carla simulator with way-points as inputs and throttle-break and steer as output. Vehicle is capable of navigating in chartered town maps • Model-based RL. Implemented model-based reinforcement learning system for autonomous driving, which includes uncertainty measurements, following framework from PILCO Show less In the autonomous driving group, I focused in decision making model of the vehicle using Reinforcement Learning techniques. Given the waypoints along the road ahead, I help implemented two models: • End-to-end model. Trained Trust Region Policy Optimization and Generative Adversarial Imitation Learning driving control unit in Carla simulator with way-points as inputs and throttle-break and steer as output. Vehicle is capable of navigating in chartered town maps • Model-based RL… Show more In the autonomous driving group, I focused in decision making model of the vehicle using Reinforcement Learning techniques. Given the waypoints along the road ahead, I help implemented two models: • End-to-end model. Trained Trust Region Policy Optimization and Generative Adversarial Imitation Learning driving control unit in Carla simulator with way-points as inputs and throttle-break and steer as output. Vehicle is capable of navigating in chartered town maps • Model-based RL. Implemented model-based reinforcement learning system for autonomous driving, which includes uncertainty measurements, following framework from PILCO Show less



OOCL

• Hong Kong
• Maritime Transportation
• 700 & Above Employee

• Data Science Intern

  • Jun 2018 - Sep 2018

  OOCL is the one of the largest container shipping company in the world. The large shipping capacity necessarily requires a huge Information Service department that manages all IT infrastructure and a data science team to optimize container scheduling and routing. I was fortunate to have witnessed the application of Machine Learning techniques in the industry. My duties involved. • Built up the entire pipeline from data gathering to performance testing from scratch • Predicted… Show more OOCL is the one of the largest container shipping company in the world. The large shipping capacity necessarily requires a huge Information Service department that manages all IT infrastructure and a data science team to optimize container scheduling and routing. I was fortunate to have witnessed the application of Machine Learning techniques in the industry. My duties involved. • Built up the entire pipeline from data gathering to performance testing from scratch • Predicted Empty-Container Daily Release Returns quantity for ports across the world with ARIMA, LightGBM and Seq2Seq model respectively. Model performance surpasses that developed by MSRA for Long Beach Port • Imputed Vessel Utility and Empty Container Re-stowage with boosting, attained performance gain upon existing implementation Show less OOCL is the one of the largest container shipping company in the world. The large shipping capacity necessarily requires a huge Information Service department that manages all IT infrastructure and a data science team to optimize container scheduling and routing. I was fortunate to have witnessed the application of Machine Learning techniques in the industry. My duties involved. • Built up the entire pipeline from data gathering to performance testing from scratch • Predicted… Show more OOCL is the one of the largest container shipping company in the world. The large shipping capacity necessarily requires a huge Information Service department that manages all IT infrastructure and a data science team to optimize container scheduling and routing. I was fortunate to have witnessed the application of Machine Learning techniques in the industry. My duties involved. • Built up the entire pipeline from data gathering to performance testing from scratch • Predicted Empty-Container Daily Release Returns quantity for ports across the world with ARIMA, LightGBM and Seq2Seq model respectively. Model performance surpasses that developed by MSRA for Long Beach Port • Imputed Vessel Utility and Empty Container Re-stowage with boosting, attained performance gain upon existing implementation Show less



PwC

• Professional Services
• 700 & Above Employee

• Associate Trainee

  • Jul 2017 - Aug 2017

  As a first year undergraduate, I was thrilled to witness the meticulousness that associates held towards documents in the PwC HK Tax Department, especially during the peak season for tax reporting. My duties involved: • Drafted Profits Tax Return, Tax Calculation, client letter for over ten clients; drafted IRD (Inland Revenue Office) letters and answered queries from IRD • Translated client’s Master File for Transfer Pricing compliance for an ICO • Automated scanning process… Show more As a first year undergraduate, I was thrilled to witness the meticulousness that associates held towards documents in the PwC HK Tax Department, especially during the peak season for tax reporting. My duties involved: • Drafted Profits Tax Return, Tax Calculation, client letter for over ten clients; drafted IRD (Inland Revenue Office) letters and answered queries from IRD • Translated client’s Master File for Transfer Pricing compliance for an ICO • Automated scanning process with Windows Batch Scripting Show less As a first year undergraduate, I was thrilled to witness the meticulousness that associates held towards documents in the PwC HK Tax Department, especially during the peak season for tax reporting. My duties involved: • Drafted Profits Tax Return, Tax Calculation, client letter for over ten clients; drafted IRD (Inland Revenue Office) letters and answered queries from IRD • Translated client’s Master File for Transfer Pricing compliance for an ICO • Automated scanning process… Show more As a first year undergraduate, I was thrilled to witness the meticulousness that associates held towards documents in the PwC HK Tax Department, especially during the peak season for tax reporting. My duties involved: • Drafted Profits Tax Return, Tax Calculation, client letter for over ten clients; drafted IRD (Inland Revenue Office) letters and answered queries from IRD • Translated client’s Master File for Transfer Pricing compliance for an ICO • Automated scanning process with Windows Batch Scripting Show less