Curriculum Vitae Web

Role: Solo Developer Timeline: 2024–Present Platform: Web (Cloudflare Pages + Render)

A full-stack web application that replaces the original desktop-only CV Manager with a multi-user, browser-based platform. Clinical trial teams can manage research experience across multiple sites, process documents, and collaborate in real time — all with role-based access control and a complete audit trail.

View Source Code Live Demo (Internal)

The Problem

The original desktop CV Manager worked well for individual users, but clinical research organizations needed a collaborative solution:

Multiple coordinators needed simultaneous access to the same study database.
Admins needed to control which users could view or edit each clinical site.
Document processing (Update/Inject, Redact) had to work in a browser without installing Python.
All changes needed to be audit-logged for compliance.
The system needed to be accessible from any device without client installation.

Solution Overview

Multi-User Sites & Studies

Per-site study databases with full CRUD operations
Import/export between Excel and the web database
Search, filter, and category-based grouping
Compressed study descriptions to minimize database footprint
Per-user category ordering and display preferences

Document Processing

Browser-based Update/Inject — uploads .docx and .xlsx, returns processed document
Redact Protocols — server-side protocol detection and masking
Parses Word documents using the same proven logic as the desktop app
Preserves all original formatting in the returned .docx

Access Control & Security

JWT-based authentication with access and refresh tokens
PBKDF2-SHA256 password hashing with brute-force rate limiting
Role-based tabs: Admin, Editor, Viewer with per-view permissions
Site-level access: admins grant view or edit per user per site
Global IP rate limiting and per-user processing rate limits

Admin & Operations

User CRUD, role assignment, and password reset
Live presence bar showing active users (admin-only)
Audit trail of every write action with user, timestamp, and changed values
Rotating auto-backups plus on-demand manual backup/restore
Forward-only schema migrations applied automatically on startup

Tech Stack

Python 3.11+ FastAPI psycopg (async) PostgreSQL (Neon) React 18 TypeScript Vite TailwindCSS Zustand Cloudflare Pages Render

Screenshots

Web application showing the Sites Dashboard with a list of clinical sites and study counts — Sites Dashboard — list of clinical sites with study counts and search

Web application showing the Study Browser with studies grouped by phase and subcategory — Study Browser — studies grouped by phase and subcategory with inline editing

Web application showing the Update and Inject document processing view — Update and Inject — server-side document processing with file upload and preview

Web application showing the Redact Protocols view for masking sensitive study information — Redact Protocols — protocol number masking for blinded regulatory submissions

Challenges & Solutions

Server-side document processing: The desktop app's python-docx logic had to run safely on the server. I isolated the processing layer in FastAPI endpoints with strict input validation, rate limiting, and temporary file cleanup.
Compressed study storage: Study descriptions can be lengthy. I implemented zlib compression on description fields before database storage, cutting row size significantly without sacrificing query performance.
Role-based tab routing: Different roles see different views. I built a TabRoute component that checks role permissions and redirects unauthorized users to their first allowed view.
JWT refresh without flashing: Early attempts caused 401 → logout cascades. I carefully sequenced the theme sync and auth refresh effects so the app validates tokens before any authenticated API calls.
Stateless backend design: Every request is fully stateless — no server-side sessions beyond JWT validation. This simplifies horizontal scaling and keeps the deployment on Render's free tier viable.

Outcomes & Impact

Migrated a solo desktop tool into a collaborative web platform with zero client installation.
Reduced per-CV processing time from 30–60 minutes to under 2 minutes via server-side document automation.
Enabled real-time multi-user access to study databases with granular site-level permissions.
Complete audit trail satisfies compliance requirements for clinical research environments.
Successfully deployed for internal use with Cloudflare Pages (frontend) and Render (backend).

Lessons Learned

Reusing the desktop app's parsing and processing logic saved weeks of development, but required careful sandboxing for server-side execution.
Async PostgreSQL with psycopg v3 and connection pooling is essential for FastAPI — naive synchronous DB access quickly bottlenecks under concurrent users.
JWT refresh logic in SPAs is subtle; the order of effect hooks and API interceptors matters enormously for avoiding logout cascades on app startup.

Future Improvements

Batch document processing for multiple CVs at once.
In-browser diff viewer comparing uploaded and processed documents.
Email notifications for pending protocol redactions or audit alerts.
Dark mode persistence synced across devices via user preferences API.