- Led team that automated and digitized a equipment installation process by a 3x speedup for fortune 500 company- Built a hair-type classification Microservice with 80% accuracy using a Residual Network trained on CelebA- Built a smart contract staking solution for Solana-based game company that commoditized in-game assets

Part of Tsinghua Statistical AI Learning Group- Experimented with segmentation object-centric deep generative model to detect adversarial data outside its distribution.- Reproduced papers including Generative Well-Intentioned Networks and Neural Independent Component Estimation.

Developed physically augmented convolutional network classifiers with 93% accuracy on earthquakes, fusing seismic and infrasonic signalsPublication"Improved discrimination of explosion signals via joint deep learning of seismic and infrasonic phases"https://spie.org/SI/conferencedetails/chemical-biological-radiological-nuclear-explosives-cbrne?SSO=1

Developed an algorithm for dynamic biological network prediction with at least 80% F1 and balanced accuracy scores on real data. Publication"Alignment-based Network Construction Algorithm"https://ieeexplore.ieee.org/document/8738865/https://dl.acm.org/citation.cfm?id=3233576

- Built back-end production architecture that integrated with a third-party vendor to verify the identity of users.- Saved $100k/year, implementing an automated user verification system

- Extended a mobile testing unit to iOS with Appium, improved runnable devices from 1 to 3.- Implemented generator integration to access a new web menu in automated testing framework.