Jimi Xia

Home

About

Projects

About Me

Hey, I’m Jimi! I graduated from RPI with a B.S. in Computer Science (GPA: 3.62) and I’m currently working as a Machine Learning Engineer at MiRABO Biotechnology in Shanghai, where I build predictive models for antibody drug development using tools like XGBoost, LightGBM, and ESM-2 embeddings.

Previously, I interned at EMAST Corp. where I built Python and C++ tools for mass spectrometry data processing, and I contributed to open-source cloud security projects at the Rensselaer Center for Open Source. I’ve also taught CS courses at Queens College, designing lab assignments and automating grading for 150+ students.

I’m passionate about full-stack development, machine learning, and building tools that solve real problems. My project work spans from webcam-driven physical therapy games with TensorFlow.js to interactive NYC mapping applications. When I’m not coding, you’ll find me hanging out with my cat Ponyo, playing basketball, or working on personal projects. Thanks for stopping by!

Experience

Machine Learning Engineer

MiRABO Biotechnology · Shanghai, CN · Jan 2026 – Present

Developed hybrid regression models achieving 0.737 Spearman correlation across 379 samples by combining ESM-2 embeddings, physicochemical descriptors, and CDR-level feature engineering. Built ensemble ML models achieving 0.823 Spearman correlation using sequence-based feature extraction and cross-validation pipelines.

Computer Science Instructor

Queens College · Queens, NY · Jun 2025 – Aug 2025

Designed Python and C++ lab assignments covering algorithms, data structures, and problem-solving for 150+ students. Automated grading workflows using Makefiles and Bash scripts, reducing evaluation time by 75%.

Software Engineer Intern

EMAST Corp. · Auburn, CA · Jun 2024 – Aug 2024

Increased data processing throughput by 20% by building Python-based analysis and automation tools for scientific instrument workflows. Developed C++ software for real-time serial communication with laboratory hardware, reducing workflow delays by 30%.