Skip to content
View Evihut's full-sized avatar

Block or report Evihut

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Evihut/README.md

Zhengji Zhang

Computer Science & Mathematics at the University of Wisconsin–Madison

I build machine learning systems and applied modeling tools with an emphasis on reproducible evaluation, clear system boundaries, and measured results.

Selected projects

Inference engineering extensions for MiniMind, including static KV caching, dynamic request batching, and observable FastAPI serving. Reproducible CPU microbenchmarks compare cache and batching configurations while documenting their experimental limits.

PyTorch · FastAPI · Static KV Cache · Dynamic Batching · Prometheus

A leakage-aware modeling benchmark, SHAP evidence bundle, and validated reporting API. A fixed-parameter evaluation records CatBoost OOF RMSE 4.870 against a training-fold mean baseline of 6.890 across 430 eligible companies.

CatBoost · SHAP · scikit-learn · FastAPI

2025 Zhixiang Cup · Grand Prize
The award applies to the original competition case; the repository documents the subsequent engineering release.

Interests

Machine learning systems · Applied mathematics · Interpretable modeling · Quantitative research

View all repositories →

Pinned Loading

  1. esg-risk-reporting esg-risk-reporting Public

    Explainable ESG risk modeling and reporting: reproducible CatBoost baselines, SHAP evidence, and a FastAPI service. 2025 Zhixiang Cup Grand Prize project.

    Python 1

  2. minimind-systems minimind-systems Public

    Personal ML systems practice based on MiniMind: static KV cache, reproducible CPU benchmarks, batched FastAPI serving and Prometheus.

    Python