PreDAP
AI-powered PreDAP simplifies onboarding, cuts repetition, and keeps your data private with real-time screen-aware intelligence.
Created on 5th April 2025
•
PreDAP
AI-powered PreDAP simplifies onboarding, cuts repetition, and keeps your data private with real-time screen-aware intelligence.
The problem PreDAP solves
PreDAP is a privacy-first Chrome Extension that automates repetitive onboarding tasks and guides users step-by-step through complex digital workflows (like setting up CRMs, configuring security settings, etc.).
What it solves:
-
Most corporate workflows (e.g., onboarding into tools like Salesforce, Okta) are filled with manual steps, often confusing for new users.
-
Companies usually rely on static docs or 1-on-1 help, which is time-consuming, inconsistent, and not scalable.
-
Our tool reduces dependency on human trainers and helps even non-tech users complete technical setups easily.
How it helps:
-
Uses AI-generated guides (like "Click [Login_Button]") to walk users through each screen interactively.
-
Automatically captures screen actions and builds a memory of tasks, so users never have to repeat manual setups.
-
Ensures complete privacy screenshots and sensitive data are abstracted and never sent to the cloud unless the user agrees.
-
Great for corporate onboarding, security configurations, or even training remote teams without risking privacy.
In short, PreDAP = Private + Productive + Personalized Automation.
Challenges we ran into
Like any innovative build, PreDAP came with its own share of roadblocks. Here are the key challenges we tackled:
Initial Pivot – Android Permissions Issue
We originally aimed to help elderly users navigate mobile UIs with an overlay-based Android assistant. However, Android’s strict permission restrictions (like Draw Over Apps and Accessibility Service) caused constant crashes and limitations. So, we pivoted to a browser-based Chrome extension for better flexibility and control.
Privacy-first AI – Sensitive Info Hiding
One of the toughest challenges in our current solution was maintaining user privacy. We couldn't directly send raw screenshots to the AI model. To solve this, we planned an Abstracter module that replaces sensitive data (like passwords, emails) with abstract labels before AI processing.
Element Position Matching – AI Alignment Issue
Another hurdle was syncing the UI element positions with the step-by-step guide generated by AI. Many HTML tags didn’t match the expected element types (e.g., buttons inside
divs
, or non-interactive labels), so we had to implement a custom mapping logic to ensure precise identification.Persistent Screenshot Capture – Chrome Extension Limitation
Chrome extensions kill scripts on page refreshes, which broke our attempt to take screenshots after each substep. To fix this, we restructured our background scripts using Offscreen Documents API and added complex tab-tracking logic to maintain continuity across steps and pages.
Each challenge helped us improve and refine the solution making PreDAP more secure, more stable, and more practical for real-world onboarding and automation.
Tracks Applied (3)
Gemini API
Major League Hacking
MongoDB Atlas
Major League Hacking
Track: GitHub
GitHub
Technologies used