Skip to content
View billzi2016's full-sized avatar
🎯
Focusing
🎯
Focusing
8 followers · 20 following

Block or report billzi2016

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.

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
billzi2016/README.md

Bill Zi

CS Ph.D., Purdue University, Indiana University

LLMs/MLLMs | AI Agentic Systems | Agent Orchestration & Harnesses | RAG | NLP & CV | GenAI | Industrial & MedAI | XAI & Robustness | mmWave & Embedded

Builder of compilers, LLM systems, distributed labs, embedded prototypes, and from-scratch engineering projects.

Homepage: billzi2016.github.io · GitHub Project · Docs
Google Scholar: Ziqian Bi

Featured Projects

  • AI-Solar-Panel-Defect-Detection-Deployment · Docs
    Solar-cell and photovoltaic-panel defect detection workspace with dataset preparation, YOLO detection experiments, anomaly detection, deployment helpers, and bilingual MkDocs documentation.

  • Gomoku-AI · Demo · Docs
    Browser-based Rust/Wasm Gomoku AI with bitboards, VCF search, alpha-beta pruning, iterative deepening, Web Worker parallelism, and local CPU execution.

  • LLM-Benchmark-Studio
    Local-first LLM evaluation platform with Django, Vue, PostgreSQL, RabbitMQ, Celery, SSE, and multi-provider model execution.

  • Build-DeepSeek-Step-by-Step
    Notebook-first deep dive into modern LLM architecture, covering tokenization, attention, GQA, MLA, MoE, training, and alignment.

  • Apple-Silicon-LLM-Engine-from-Scratch
    GPT-2 inference engine rebuilt across NumPy, PyTorch MPS, and C++/Metal stages for Apple Silicon.

  • distributed-paxos-raft-lab
    Distributed consensus lab with FastAPI node simulation, Vue dashboards, Docker orchestration, and cluster control workflows.

  • GPU-Test-and-Polite-Scheduler
    Practical shared-GPU toolkit for environment checks, stress tests, communication diagnostics, and tmux-based polite scheduling.

  • Leetcode-All-Languages-Best-Solutions · Docs
    Multi-language LeetCode solution workspace with generated Markdown outputs, difficulty/range organization, and a MkDocs documentation site.
    An AI agent that controls reasoning level to save tokens while still producing accurate answers, using the large language model (LLM) gpt-oss 120b through an Ollama host.
    Languages:
    C, C++, Java, Python, Python3, C#, JavaScript
    TypeScript, PHP, Swift, Kotlin, Dart, Go, Ruby
    Scala, Rust, Racket, Erlang, Elixir.

  • homemade-risc-v-64-vector-linux-emulator · Docs
    From-scratch educational 64-bit RISC-V emulator with Vector extensions and Linux boot support.

  • Industrial-Query-Agent
    Industrial query agent project for structured domain question answering, retrieval-oriented workflows, and practical agentic system design.

  • Othello-AI · Demo · Docs
    Browser-based Rust/Wasm Othello AI with bitboards, alpha-beta search, exact endgame search, stability-aware evaluation, and Web Worker parallelism.

  • automl-autodl · Docs
    Unified tabular AutoML and AutoDL workspace for Home Credit risk modeling, with shared preprocessing, classical ML, PyTorch models, and documentation pages.

  • C-Core-Compiler · Docs
    C-like compiler project covering lexical analysis, parsing, semantic checks, IR, code generation, and documentation.

  • python-git-reproduction · Docs
    Reproducible Python and Git workflow documentation project with MkDocs pages and repository-level maintenance notes.

  • DeepChrInteract-v2 · Docs
    Deep learning project for chromatin interaction modeling with a Sphinx documentation site.

  • MIMO-FMCW-Radar-Simulator-Multiprocess · Docs
    Multiprocess MIMO FMCW radar simulation project with public documentation pages.

  • mmwave-fmcw-cascade-mimo-sensing-platform · Docs
    mmWave FMCW cascade MIMO sensing platform with documentation for radar signal processing and system workflows.

  • mmlock-fmcw-radar-deep-security · Docs
    FMCW radar deep-security project for mmWave sensing, authentication, and documentation-driven maintenance.

  • whisper-meeting-transcription-translation-and-summary
    Offline subtitle generation and bilingual translation tool for local audio and video files, accelerated with MLX on Apple Silicon and powered by local LLM workflows.

Systems, Compilers, and Runtime

LLM, Agents, and Evaluation

Research, Data, and Domain Systems

Infra, Observability, and Tooling

Embedded, Hardware, and Edge

Algorithms, Search, and Game AI

Practice and Study Repositories

More

Scholar

Google Scholar Profile

Google Scholar profile snapshot

Recent GitHub Pages Maintenance

This maintenance work turns experimental, research-oriented, and long-running repositories into clearer public-facing projects. It includes reorganizing repository structures, rewriting and standardizing README files, establishing consistent project entry points, and adding practical explanations for installation, execution, architecture, datasets, experiments, and expected outputs. English and Chinese documentation are kept aligned where bilingual presentation is useful.

The documentation layer has also been strengthened across repositories through clearer navigation, cross-page links, mobile-friendly layouts, stable asset paths, and GitHub Pages deployment. Depending on the needs of each project, the public-facing documentation now converges around three delivery styles: MkDocs documentation sites, Sphinx documentation sites, and fully custom static sites built with HTML, CSS, and JavaScript.

The maintained portfolio spans compilers and runtime systems, LLM and agent engineering, computer vision and medical imaging, radar simulation and mmWave sensing, reproducible research workflows, and browser-based game AI such as Gomoku and Othello. The goal is to make each repository easier to inspect, understand, reproduce, and maintain over time, even for visitors who have not yet run the project locally. The examples below show several representative documentation and presentation patterns used in this ongoing cleanup.

MkDocs documentation site and repository README / documentation system: python-git-reproduction

MkDocs documentation site for python-git-reproduction Repository README and documentation system for python-git-reproduction

Sphinx documentation site: DeepChrInteract-v2

Sphinx documentation site for DeepChrInteract-v2

Static HTML/CSS/JS site: billzi2016.github.io

Static personal site billzi2016.github.io

Curated project README example: Awesome Flow Matching

Curated project README example from Awesome Flow Matching

AI-Assisted Maintenance Activity

Recently, I have been refactoring and polishing my GitHub Pages presence, using Codex / Claude Code within a Human-in-the-Loop workflow to turn older repositories into more formal, better-documented, and more navigable project sites. The screenshots below summarize the past month's usage and maintenance activity.

The work is not about blindly generating pages. It follows a Spec-First and Review-Driven process, with Test-Driven Development (TDD), Spec-Driven Development (SDD), and Continuous Integration / Continuous Delivery (CI/CD) practices used to improve readability, cross-platform support, maintainability, long-term maintenance efficiency, security, and stability across legacy repositories. AI assistance greatly accelerates the overall pace of restructuring and cleanup, and when dealing with large, interdependent documentation sets, it has a natural advantage in structural organization, terminology alignment, and cross-document revision, which helps reduce the risk of updating one section while leaving related documentation behind. At the same time, project decisions, acceptance, consolidation, and final editorial control remain Human-in-the-Loop.

Cross-model token consumption Codex usage dashboard showing 18,935,155,313 tokens consumed across models

Conversation turns by model Codex usage dashboard showing 61,090 conversation turns by model

Lines of code generated Codex usage dashboard showing 3,429,823 lines of code generated over time

Pinned Loading

  1. ClinicaLLM-OmniBench-EN ClinicaLLM-OmniBench-EN Public

    Lightweight medical benchmark framework for evaluating LLMs on exam-style and specialty question banks with standardized pipelines and resumable runs.

    Python 1

  2. RS485-Modbus-Concrete-Sensor-Monitor RS485-Modbus-Concrete-Sensor-Monitor Public

    混凝土传感器 RS485/Modbus TCP 实时监测系统，基于 Django + Redis，支持真实数据/网络模拟/本地模拟三档切换

    Python

  3. DeepChrInteract-v2 DeepChrInteract-v2 Public

    PyTorch reimplementation of DeepChrInteract: EPI prediction with 14 encoder architectures (CNN, BiLSTM, Transformer, Mamba, RWKV, mLSTM, MAE, DNA LLMs)

    Python

  4. mmlock-fmcw-radar-deep-security mmlock-fmcw-radar-deep-security Public

    mmWave radar security project focused on user-away detection, anti-theft protection, radar imaging, and deep-learning-based sensing workflows.

    Jupyter Notebook

  5. llm-auto-rag-survey llm-auto-rag-survey Public

    Literature discovery pipeline for LLM, RAG, agent, multimodal, and evaluation papers, built to organize arXiv research into structured survey workflows.

    Python

  6. MRI_Deep_Learning_Projects MRI_Deep_Learning_Projects Public

    Collection of MRI deep-learning projects spanning public neuroimaging datasets, supervised tasks, self-supervised pretraining, and interpretability analysis.

    Python