Group Project

HaloAudit

AI-powered document audit system for scanning files for sensitive data, security risks, and quality issues.

HaloAudit is a HackHarvard 2025 winning group project, selected against 180 competing teams. It combines desktop drag-and-drop apps, including a macOS app and Windows port, a Cloudflare Workers edge backend, and a Python agent pipeline to audit PDFs, CSVs, and spreadsheets and return clear, actionable reports.

GitHub
HaloAudit
Project Preview
HaloAudit audit workflow visualHaloAudit audit workflow
My Role

As part of the group project, my visible GitHub contributions under TylersHub were focused on the Windows frontend port, project documentation, and integration/presentation work across the audit workflow. I present HaloAudit as a collaborative build and do not claim sole ownership of the full system.

  • Added and integrated the Windows frontend port for the HaloAudit desktop audit experience
  • Worked on Windows-side upload, findings, API client, WebSocket, and audit view model files
  • Updated project documentation and README content to explain the architecture, setup, and audit pipeline
  • Merged and coordinated integration changes across the desktop clients, backend API, Python agent, and report flow
  • Helped present the system as an end-to-end document audit workflow for reviewers and users
Problem

Teams that handle documents often need to identify sensitive data, secrets, inconsistencies, and quality issues quickly, but manual review is slow and easy to miss.

Solution

HaloAudit provides a drag-and-drop audit workflow where files are uploaded, queued, processed by an AI-assisted agent pipeline, and returned as readable audit reports with progress updates.

Impact

Selected as one of the winning projects at HackHarvard 2025 against 180 competing teams, HaloAudit demonstrates a multi-part product architecture across desktop UI, a Windows port, real-time status updates, cloud edge services, storage, database queues, vector indexing, and AI-powered document analysis.

Key Features
  • Desktop drag-and-drop file upload experience for PDFs, CSVs, and spreadsheets, with a macOS app and Windows port
  • Real-time progress updates through WebSocket-based status handling
  • Cloudflare Workers backend with signed R2 uploads, D1-backed job queue, Durable Objects, and Vectorize integration
  • Python LangGraph agent pipeline for ingesting, extracting, chunking, embedding, checking, analyzing, and reporting
  • AI-powered report generation for sensitive data, security risks, and document quality issues
Tech Stack
  • Swift
  • SwiftUI
  • Windows
  • TypeScript
  • Cloudflare Workers
  • Hono
  • D1
  • R2
  • Durable Objects
  • WebSockets
  • Vectorize
  • Python
  • LangGraph
  • Gemini
  • Next.js