Infillo AI

AI-powered form autofill that learns from your documents and context to fill forms intelligently across web and browser.

Comment

Inspiration

The inspiration for Infillo AI came from the frustration of repeatedly filling out lengthy forms with the same information, whether for job applications, insurance claims, or government services. While Google Autofill exists, it's limited to basic field matching and doesn't understand context or provide intelligent suggestions. We envisioned an AI-powered assistant that could understand form semantics, learn from user documents, and provide contextual, personalized suggestions that go far beyond simple text matching.

What it does

Infillo AI is an intelligent form-filling assistant that revolutionizes how users interact with web forms. Unlike Google Autofill, it:

🧠 Smart Form Analysis: Uses AI to analyze HTML structure and understand field semantics, not just basic name/type matching

📄 Document Intelligence: Processes uploaded documents (PDF, DOCX, TXT) to extract entities and build a personalized knowledge base

🎯 Contextual Suggestions: Provides AI-generated suggestions based on form context, user documents, and semantic understanding

💬 Conversational AI: Offers a chat interface for form assistance and refinement of suggestions

🔄 Learning System: Improves accuracy through user feedback and form submission history

🛡️ Enterprise Security: Implements Google OAuth2, JWT authentication, and complete CSS isolation using Shadow DOM

📊 Analytics Dashboard: Provides insights into form usage, accuracy statistics, and suggestion performance

How we built it

Architecture: Built as a modern monorepo with three main components:

  1. Chrome Extension (Manifest V3):

    • React + TypeScript frontend with complete CSS isolation using Shadow DOM
    • Content script for intelligent form detection and widget injection
    • Real-time communication with backend APIs

  2. Backend API (Express.js + TypeScript):

    • Google AI (Gemini) for semantic analysis and suggestion generation
    • MongoDB Atlas with vector search for semantic similarity matching
    • Google Cloud Storage for document processing and storage
    • Dynamic provider system supporting multiple AI and storage backends

  3. Web Dashboard (Next.js 14):

    • User context management and document upload interface
    • Form history and analytics visualization
    • Authentication and settings management

Key Technologies:

  • AI/ML: Google Generative AI (Gemini), vector embeddings, semantic search
  • Database: MongoDB with Atlas Vector Search (768-dimensional embeddings)
  • Storage: Google Cloud Storage with signed URLs
  • Security: OAuth2, JWT, Shadow DOM isolation, comprehensive CSP

Challenges we ran into

1. CSS Isolation: Ensuring the extension UI doesn't interfere with website styles required implementing complete Shadow DOM isolation with custom CSS reset strategies

2. Form Detection Complexity: HTML forms vary wildly across websites. We built multiple extraction strategies using JSDOM parsing, aria-label detection, and fallback regex patterns

3. AI Context Management: Balancing context richness with API token limits while maintaining response speed required sophisticated prompt engineering and context prioritization

4. Vector Search Performance: Implementing efficient semantic search required optimizing MongoDB Atlas vector indexes and handling user-specific filtering without performance degradation

5. Real-time Suggestions: Achieving sub-second response times for AI suggestions while processing complex form contexts and user documents

6. Cross-Origin Security: Implementing secure communication between extension, web pages, and backend while maintaining strict security policies

Accomplishments that we're proud of

🎨 Zero UI Interference: Achieved complete CSS isolation using Shadow DOM, ensuring the extension never breaks website layouts - a major improvement over many existing extensions

🧠 Advanced AI Integration: Successfully implemented contextual AI that understands form semantics beyond simple field matching, providing genuinely helpful suggestions

📄 Document Intelligence: Built a robust document processing pipeline that extracts meaningful entities from various file formats and makes them searchable

⚡ Performance Optimization: Achieved sub-second AI response times through efficient prompt engineering and vector search optimization

🔒 Enterprise-Grade Security: Implemented comprehensive security including OAuth2, JWT refresh tokens, rate limiting, and content security policies

🎯 High Accuracy: Achieved significantly better suggestion accuracy than basic autofill through semantic understanding and contextual learning

🔧 Extensible Architecture: Built a provider pattern allowing easy integration of different AI services (OpenAI, Anthropic) and storage backends (AWS, Azure)

What we learned

AI Prompt Engineering: Learned sophisticated techniques for context management, embedding generation, and maintaining consistency across different AI models

Browser Extension Architecture: Mastered Manifest V3 best practices, Shadow DOM implementation, and secure content script communication patterns

Vector Database Optimization: Gained deep expertise in MongoDB Atlas Vector Search, including index optimization and hybrid search strategies

User Experience Design: Discovered the importance of non-intrusive UI design and the complexity of cross-website compatibility

Performance Engineering: Learned to balance AI sophistication with real-world performance constraints, including token limits and response time requirements

Security Best Practices: Implemented comprehensive security measures including CSP, CORS, rate limiting, and secure authentication flows

What's next for Infillo AI

🌐 Enhanced Form Types: Support for complex form types including file uploads, multi-step forms, and dynamic field generation

🎯 Smart Templates: Auto-generate form templates from user behavior and provide one-click form completion for common scenarios

🔗 Third-Party Integrations: Connect with CRM systems, HR platforms, and other business tools for enterprise workflow automation

📊 Advanced Analytics: Implement predictive analytics to suggest form optimizations for website owners

🌍 Multi-Language Support: Expand beyond English with localized form detection and suggestion generation

🔐 Privacy-First Options: Implement on-device AI processing options for users requiring maximum privacy

Built With

Share this project:

Updates