Bryce Tu Chi Email & Phone Number

Livermore, California, United States

•Developed a machine learning-driven multi-scale modeling infrastructure in a team of 6 engineers, automating Ultra Coarse-Grained (UCG) to Martini Coarse-Grained (CG) model transition, improving workflow efficiency by 10%• Refactored, optimized, and tested codebase from Perl to Python; automated the CG-to-UCG transition using Pandas, SciPy, Shell scripting, and MDAnalysis, reducing execution time by 25%, memory usage by 15%, and bug occurrence by 40%• Integrated MLP neural networks into the UCG-to-CG transition process using PyTorch, TensorFlow, and CUDA, reducing prediction errors by 30% and improving detail structural resolution of by 10%

Claremont, California, United States

• Designed and implemented Martian terrain classification algorithms using similarity metrics such as SSIM, GMSD, LPIPS, EMD, PSNR, and NCC, targeting 80% classification accuracy• Generated 10,000+ synthetic images leveraging Stable Diffusion, PyTorch, Hugging Face, and OpenAI API, aiming for a 40% increase in feature diversity across terrain categories• Analyzed metric sensitivity and image diversity over 5,000+ iterations, using PyTorch, OpenCV, TensorFlow, Beautiful Soup, NumPy, Pandas, Matplotlib, and Lambda Cloud, identifying key metrics that improved classification accuracy by 15%• Integrated real and synthetic images for self-supervised classification and segmentation model tuning, targeting a 10% improvement in Mars landscape analysis precision

Claremont, California, United States

• Worked with a team of 10 engineers to develop a drone-assisted bridge inspection simulation in C# and Unity, integrating real-time controller inputs, obstacle detection, velocity monitoring, and biometric sensor data synchronization• Implemented multi-threaded processing in C# to improve real-time simulation, debugging and resolving threading issues with JetBrains Rider, Visual Studio Debugger, and WinDbg, reducing simulation latency by 30%• Designed, implemented, troubleshot, and tested a real-time data processing pipeline in C#, optimizing the integration of multiple biometric sensors (eye-trackers, EEG, controllers), improving data throughput by 20%• Built convolutional neural networks and ensemble models in PyTorch, achieving 80% accuracy in inspection error prediction• Optimized model inputs using R, Python, and MATLAB, reducing training time by 20% and improving predictive performance by 15% from analysis of 300+ biometric variables for 40+ user participants

Claremont, California, United States

Claremont, California, United States

• Developed interactive web applets using Figma, HTML, CSS, and JavaScript. Optimized simulations using asynchronous JavaScript, lazy loading, and efficient DOM manipulation, reducing simulation latency by 12%• Boosted student assessment scores in Theoretical Mechanics by 20% with dynamic applets and real-time data visualizations