>Arshia Pessaran

• ├── about/
  │ └──

  Biology degree from UCF and went on to go to medical school at FIU. Withdrew to pivot into tech. Finished a second degree in data science in the span of a year and working as a software engineer ever since.

• ├── connect/
  • │ ├── github
  • │ ├── linkedin
  • │ ├── email
  • │ └── resume
• └── projects/
  • ├── Bludly
    │ └──

    A HIPAA-compliant fullstack health data platform for uploading, tracking, and analyzing blood lab results over time. Interactive D3.js visualizations chart biomarker trends with color-coded reference ranges, date filtering, and multi-lab overlay. AI analysis runs on a self-hosted Deepseek R1 model — no patient data leaves the infrastructure — generating health grades, biological age estimates, and actionable insights.
    
    Supports multiple user profiles for family management, secure time-limited doctor sharing via tokenized links, and data export. Authenticated with Google OAuth 2.0 and deployed on AWS EC2.
    
    React · Node.js · Express · PostgreSQL · D3.js · Google OAuth 2.0 · Passport.js · REST API · Docker · AWS EC2 · Deepseek R1 · Vite · Tailwind CSS
  • ├── PDF Reconstruction Pipeline
    │ └──

    A multi-stage pipeline that reconstructs scanned PDFs into clean digital documents. Preprocessing corrects page skew via projection profile analysis and Otsu thresholding, denoises with a bilateral filter to preserve edges, and binarizes for stroke analysis. Pages are classified in two passes — a fast pixel statistics pass filters obvious outliers, while a full pass combines Surya's deep learning layout model with OCR confidence metrics to label each page as normal, mixed, or outlier.
    
    A two-phase calibration system processes the first batch of pages to establish a font size and style baseline before full reconstruction. OCR is batched across pages for GPU efficiency, with support for three engines: Surya, Tesseract, and PaddleOCR.
    
    Python · PyMuPDF · ReportLab · OpenCV · Surya · PaddleOCR · Tesseract · deep learning · OCR · computer vision · image processing · pipeline architecture · customtkinter
  • └── TradingView Heatmap Bot
    └──

    A Windows background application that automates hourly capture of the TradingView S&P 500 heatmap and delivers it to a Discord channel. Uses Playwright to launch a headless Chromium browser, navigate to TradingView, configure the heatmap view, and capture a screenshot — all without user interaction. A time-aligned scheduler fires on the hour, and the captured image is posted via Discord's incoming webhook REST API.
    
    Packaged as a standalone Windows executable with PyInstaller. Runs as a persistent system tray process with pause, resume, and manual trigger controls.
    
    Python · Playwright · headless browser automation · Discord API · webhook integration · task scheduling · PyInstaller · Windows system tray