Our mascot!

Wizdom: Document Simplification, Translation, and Form Filling

Inspiration

As first-generation college students from immigrant families, we grew up helping our parents read and understand complex documents, such as medical forms, tax returns, and leases. We saw how confusing official language can be, especially for non-native English speakers. That experience inspired us to build Wizdom, a tool that helps people make sense of complicated documents and actually complete them. Our goal is to remove both the comprehension barrier and the form-filling friction.

What it does

Wizdom takes difficult documents such as legal, academic, medical, or policy texts and rewrites them in short, natural language summaries that anyone can understand. Users can paste text or upload files, and the AI generates a simplified version that keeps all key details while removing unnecessary jargon, then narrates it back to the user in their chosen language.

Beyond simplification, Wizdom extracts fillable form fields from PDFs and presents them as an intuitive form interface. Users can fill out forms with confidence, seeing field labels in their preferred language while the system generates a professionally formatted, completed PDF ready for submission. The goal is to make information accessible and empower users to understand and complete official documents without needing expert help.

How we built it

We built Wizdom using Python for the backend and Google's Gemini 2.5 Flash API for generating simplified summaries and intelligent translations. The backend extracts text efficiently and sends carefully engineered prompts to the model to produce clear, accurate results.

For form handling, we used pypdf to extract AcroForm field metadata from PDFs and ReportLab to generate professional filled PDFs with custom styling and multi-language field name support.

We used React for the frontend to create a responsive and intuitive user interface that supports text input, file uploads, live simplification, and dynamic form field rendering. To make the experience more interactive and inclusive, we integrated ElevenLabs' text-to-speech API, which allows users to listen to simplified summaries in natural, human-like voices across multiple languages. This feature is especially helpful for users who prefer auditory learning or have reading difficulties.

Challenges we ran into

One of the main challenges was finding the right balance between clarity and completeness during simplification. Early versions of the model sometimes oversimplified the text or removed important context. We solved this by refining our prompts and testing across different document types. For form filling, we encountered technical challenges with PDF manipulation. Different libraries had limitations with AcroForm reconstruction. We solved this by generating fresh, professionally formatted PDFs from scratch rather than modifying originals, which proved more reliable and user friendly. We also handled large input sizes efficiently, managed formatting issues when combining AI generated text with audio output, and optimized API calls for performance.

Accomplishments that we're proud of

We're proud that Wizdom works as a complete product, not just a prototype. It can take a dense academic policy or legal contract and transform it into something a middle school student could understand, then extract and guide users through form completion with confidence.

We successfully integrated two major AI APIs (Gemini and ElevenLabs) seamlessly, implemented dynamic form field extraction and PDF generation, and built multi language support throughout the entire workflow from simplification to form field labels.

Most importantly, this project reflects our backgrounds and values. It's meaningful to build something that could genuinely help our families and communities navigate bureaucratic systems that often exclude non-native English speakers.

What we learned

We learned how to design effective AI prompts that produce consistent results and how to balance simplicity with accuracy. We gained hands-on experience integrating multiple APIs, handling asynchronous operations, and building for accessibility with speech features.

We also learned crucial lessons about PDF manipulation. When to fight the technology and when to work around it. Sometimes the best solution isn't preserving the original but creating something better and more reliable from scratch.

We understood the importance of UX design when building tools meant for diverse audiences and real-world impact, and how language localization extends beyond simple translation.

What's next for Wizdom

Enhanced Form Intelligence: We plan to include intelligent data validation and context-aware suggestions so the tool can detect missing or inconsistent information before submission. This would guide users through form completion with confidence, flagging potential errors proactively.

Conversational Interface: We want to make Wizdom more interactive by introducing audio input alongside text. Users could speak their answers, and the system would intelligently populate form fields. More ambitiously, once a document is simplified and translated, we envision a back-and-forth conversational experience as an option to explore. Like talking with an expert over the phone who understands both the document and your needs. This would transform form-filling from a solitary task into a guided dialogue.

Advanced Document Handling: Currently, we support AcroForm PDFs. Our next goal is to extend form filling to any document type, including non-AcroForm PDFs, scanned documents, and image-based forms. We'd use computer vision and intelligent pattern recognition to identify fillable regions, extract context about what each field expects, and generate filled documents that match the original layout and formatting. This technical challenge excites us because it would unlock support for older documents and international forms that don't use modern PDF standards.

Batch Processing and Scalability: We plan to add batch processing capabilities so users can upload multiple documents, simplify them in bulk, and manage form-filling workflows at scale. This opens possibilities for institutional use, helping organizations support diverse populations through their own documentation processes.

Technical Stack Backend: Python, FastAPI, Gemini 2.5 Flash API, ElevenLabs TTS, pypdf, ReportLab Frontend: React, TypeScript, Tailwind CSS Infrastructure: Cloudflare Tunnels for local development sharing