Experience

RoboEye.ai

• Canada
• Software Development
• 1 - 100 Employee

• Computer Vision Engineer Intern

  • May 2023 - Present



New York University

• Higher Education
• 700 & Above Employee

• Lab Volunteer

  • Feb 2023 - Present

  • Parallelized an image verification procedure and reduced inference time by ~1 second for each query; set up server on remote machine and tested mobile application; incorporated codes to control camera and extract frames from videos.

• Lab Volunteer

  • Jan 2023 - Present

  • Exploring the sim2real problem that estimating the poses of manufacturing parts given RGB-D information under factory environments when computer-synthetic data is abundant but real data is rarely available.• Designed pipelines to extract 70k+ instances from raw data and modularized preprocessing, training and evaluating stages; trained the neural network and combined it with numerical optimization to control translation errors within 5 mm.

• Math Tutor

  • Jan 2023 - May 2023

• Lab Volunteer

  • Oct 2022 - Dec 2022

  • Utilized reinforcement learning to train a robot in simulated environment to move towards objects and decide when to pick and drop so that it can manipulate objects to the target positions.• Designed a double-level decision-action neural network model to tackle the problem of sparsity of atomic action space and thus efficiently completed the training process and able to train the basic action models in parallel.• By experimenting reward mechanism, termination condition and reactive behavior, trained the robot to learn how to bypass obstacles and avoid getting stuck, thus improving the success rate from 70+% to 90+%. Show less



Renmin University of China

• China
• Higher Education
• 700 & Above Employee

• Research Assistant

  • Mar 2022 - Aug 2022

  • Built an ensemble machine learning model to evaluate second-hand house prices based on 20+ million of transaction records and then transferred the model to predict foreclosure house prices in the auction market with new data obtained online. Both models achieved over 90% PPE15 (the proportion of test data with error less than 15%)• Others: queried SQL database; designed a pipeline to clean data and create factors; programmed a web scraper to collect foreclosure transactions from online market; wrote shell scripts to run models; modified project letter to pitch customers. Show less

• Research Assistant

  • Nov 2021 - Apr 2022

  • Designed a data pipeline to create 50+ micro and macro factors and trained a machine learning model to construct stock portfolio, and achieved the performance as equal as in the latest literature.• Wrote a web scraper to obtain commodity prices from government sources; created methods to compute industry-adjusted factors, to align accounting period with actual release dates, etc.



Everbright Securities

• China
• Investment Banking
• 200 - 300 Employee

• Financial Analyst Intern

  • Jan 2022 - Feb 2022

  • Python: backtested investment strategies for Chinese convertible bond market; automated the procedure of updating market data, computing bond index and selecting portfolio. • Microsoft Office: visualized market trends and edited weekly reports. • Python: backtested investment strategies for Chinese convertible bond market; automated the procedure of updating market data, computing bond index and selecting portfolio. • Microsoft Office: visualized market trends and edited weekly reports.