Skip to content
View xinyiklin's full-sized avatar

Block or report xinyiklin

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
xinyiklin/README.md

Hey there 👋

I'm Xinyi Lin — a full-stack software engineer building real-world tools at the intersection of software and healthcare.

I've spent the last two years inside a high-volume cardiology clinic watching clinicians fight their EHR every day, and I build software that tries to do better.


🧩 Featured Projects

🏥 CareFlow — Full-Stack EHR-Style Workflow App

A portfolio-grade EHR demo covering scheduling, clinical charting, billing, document management, and organization/facility administration. Modeled on the workflows I see in a real clinic — not generic CRUD.

  • Clinical charting with SOAP progress notes (draft / signed / unsigned states), encounter lifecycle, and a 68-code CPT catalog
  • Configurable fee schedules with payer and facility overrides
  • Patient timeline that unifies appointments, encounters, medications, and allergies into one chronological view via a shared timeline component
  • Security: SSN encryption at rest (Fernet), JWT + HTTP-only refresh cookies, CSRF protection, role-based permissions scoped per facility, and an audit log for sensitive actions
  • Typed end-to-end via OpenAPI codegen (drf-spectacular + openapi-typescript)
  • Deployed on AWS Amplify (frontend) + Render (backend), with PostgreSQL on Amazon RDS, Cloudflare R2 object storage, and Docker Compose for local dev

🔗 Live demo: careflow.xinyiklin.com — clinician app at clinician.xinyiklin.com, patient portal at patient.xinyiklin.com 💻 Code: github.com/xinyiklin/careflow

📄 RoleFit AI — Browser-Primary, Local-First Job Application Workbench

RoleFit combines resume tailoring, recruiter-style review, and application tracking in the browser while a small Electron companion manages the local server and AI providers.

  • Evidence-grounded AI workflow with prompt-injection fencing, output sanitization, offline adversarial evals, and reviewable edits instead of silent resume rewrites
  • Five explicit providers: Claude Code, Codex, and Antigravity through provider-owned CLI sessions, plus OpenAI and Claude APIs with keys encrypted locally by the companion
  • Built on the shared Typeset engine — edit the exported page directly, generate a deterministic client-side PDF, and save the structured document as a portable .resume file
  • Workday-aware job import with AI distillation, deterministic fallback extraction, duplicate checks, and a fail-closed multi-stage polish/review workflow
  • Local application tracker with table and calendar views; no RoleFit account or cloud workspace, and only the inputs required for the selected AI action leave the device

🔗 Product site: rolefit.xinyiklin.com 💻 Code: github.com/xinyiklin/rolefit-ai

📝 Typeset — Browser-Native Resume Editor & Typesetting Engine

A WYSIWYG resume editor built on a from-scratch deterministic typesetting engine — edit the rendered page directly and export a pixel-faithful PDF that the app renders itself. Runs entirely in your browser; your resume never leaves your device.

  • From-scratch layout engine (measure → line-break → blocks → layout) shared by the editor, browser print, and a client-side pdf-lib PDF emitter — the editor and the exported PDF render glyph-identically, verified by an automated font-parity suite
  • Direct editing on the engine-rendered page: structured document model, inline formatting marks, and real undo/redo with exact caret restoration
  • Strict versioned .resume file format with browser autosave, plus print-aware typography — six bundled font families, three of them metrically compatible with Times New Roman, Calibri, and Arial so page count survives a round trip through a word processor
  • Engine and editor live as npm workspace packages (@typeset/engine, @typeset/editor) inside the RoleFit AI monorepo, powering both Typeset and RoleFit from one source of truth
  • Dockerized and deployed to AWS EC2 (custom domain, HTTPS) via a path-filtered GitHub Actions CI/CD pipeline

🔗 Live app: typeset.xinyiklin.com 💻 Code: github.com/xinyiklin/rolefit-aiapps/typeset + packages/engine / packages/editor

🧰 Machine Bootstrap — Portable AI-Agent Workspace Setup

The tooling behind how I work with coding agents: one-time, idempotent seeding of agent guidance, shared skills, and a delivery workflow across a workspace of independent sibling repositories — without turning that workspace into a parent repo that silently owns every project's instructions.

  • Sibling-only ownership model — the workspace root holds no live guidance; each project owns its own AGENTS.md / CLAUDE.md after seeding and is never re-synced against the template source
  • Idempotent Node initializers that create only missing files, preserve existing ones, refuse unsafe targets (filesystem root, $HOME, Git-owned workspaces), and roll back only run-owned files on failure
  • .gitignore environment policy verified through Git itself rather than by pattern matching — ambiguous or ineffective rules stop for manual review before any write
  • Supply-chain-pinned shared skills: every skill pinned by repository, revision, and content SHA-256, with a missing-only installer and drift detection
  • Provider-neutral workflow packages (Product Partner → Delivery Lead → Verifier, with explicit approval gates) compiled into deterministic Claude and Codex agent adapters
  • Verified on Linux and Windows in CI against a disposable regression suite covering rollback, concurrency, and fail-closed paths

💻 Code: github.com/xinyiklin/machine-bootstrap


🛠 Tech Stack

Languages: Python, TypeScript, JavaScript, C++, SQL, HTML/CSS

Frontend: React, TypeScript, Vite, React Query, React Router, Tailwind CSS

Backend: Django, Django REST Framework, Node.js, REST APIs, JWT + CSRF auth

Data & Storage: PostgreSQL, Cloudflare R2 / S3-compatible object storage

Desktop & Local Runtime: Electron, loopback Node services, OS-encrypted credential storage

Tooling & Deployment: Git, Docker, npm workspaces (monorepo), AWS (Amplify, RDS, EC2), GitHub Actions CI/CD, GitHub Pages, Cloudflare, Render, ESLint, Prettier, pre-commit hooks


🎯 About Me

  • 🏥 Two years inside a cardiovascular clinic — real exposure to EHR migrations, scheduling bottlenecks, and clinical workflow design
  • 🛠 Comfortable across the full stack: schema design, REST APIs, React frontends, auth/security, deployment
  • 🌱 Currently going deeper on system design, OpenAPI tooling, and clinical informatics
  • 🌐 Bilingual in English and Mandarin Chinese

🌐 Get in Touch


Currently exploring early-career and SWE I roles in healthcare tech and full-stack — always open to a coffee chat about clinical software, full-stack architecture, or building tools clinicians actually want to use.

Pinned Loading

  1. careflow careflow Public

    Full-stack EHR-style clinic platform — scheduling, clinical charting, billing, patient management, secure messaging, and a multilingual patient portal. React + Django REST Framework + PostgreSQL.

    TypeScript 2 2

  2. xinyiklin.github.io xinyiklin.github.io Public

    Personal portfolio site (React + Vite) for full-stack engineering roles with a healthcare software focus — leads with the CareFlow case study. Live at xinyiklin.com.

    JavaScript 1

  3. rolefit-ai rolefit-ai Public

    Browser-primary, local-first resume tailoring and application tracking with an Electron provider companion, evidence-grounded AI, deterministic PDF export, and five explicit CLI/API providers.

    TypeScript 1

  4. catch-the-ball catch-the-ball Public

    A 2D arcade game built in C++ with GLFW and OpenGL — move a paddle to catch falling balls with scaling difficulty and real-time scoring. Hunter College CSCI 39541 course project.

    C

  5. machine-bootstrap machine-bootstrap Public

    Portable AI-agent workspace bootstrap and shared guidance

    JavaScript