InspirationThe seed for Presstigator was planted when we looked at the modern landscape of investigative journalism.
Today’s reporters are drowning in data but starving for clarity. Whether it is analyzing massive leaks or tracking shell companies, journalists spend roughly $70\%$ of their time on "data grubbing" , the manual, tedious process of connecting dots across scattered files.We were deeply inspired by the courage of press professionals who face physical and digital dangers. We wanted to build a "digital shield".A tool that not only automates the grunt work but also uses secure proxies to protect a journalist's identity while they scour the web for the truth.
What it doesPresstigator is an AI-powered OSINT (Open Source Intelligence) engine.
A journalist simply uploads a collection of files , photos of documents, PDFs, or spreadsheets. Ingestion: The system cleans the files, stripping dangerous EXIF metadata to protect sources.Analysis: Using Gemini 1.5 Pro, it extracts entities (People, Companies, Locations).Expansion: It automatically triggers a web-wide search via secure proxies to find public records, social media links, and news mentions.Reporting: It synthesizes everything into a professional, structured intelligence report, revealing hidden connections that would take a human weeks to find.
How we built itWe built a decoupled, full-stack architecture designed for speed and security:
Frontend: Developed with Next.js and Tailwind CSS, focusing on a "Dossier-style" UI that feels like professional forensic software.Backend: A FastAPI (Python) server handles the heavy lifting, orchestrating the data pipeline. AI Engine: We integrated the Gemini API, leveraging its massive context window to process entire folders of evidence simultaneously. Database: We utilized SQLAlchemy with a role-based access control system (Owner, Company, Journalist) to ensure data silos and security. Security: Implemented JWT (JSON Web Tokens) and encryption for sensitive document handling.
Challenges we ran intoThe primary hurdle was Time. Building a multi-role system with a complex AI pipeline from scratch in such a short window required intense coordination.
Technically, we faced significant obstacles with Proxy integration. Finding a balance between high-speed OSINT scraping and the legal/ethical boundaries of data collection was tough. We also wrestled with the cost and complexity of professional-grade scraping APIs, forcing us to optimize our query logic to get the most "intelligence" for the lowest computational cost.
Accomplishments that we're proud ofWe are incredibly proud of creating a functional, end-to-end pipeline that actually works. Seeing a raw "Capture d'écran" (screenshot) or a blurry PDF be transformed into a structured JSON entity list, and then into a formatted report, felt like magic. We successfully built a platform that balances a sophisticated aesthetic with "hard" security features like metadata stripping and role-based permissions.
What we learnedThis project was a masterclass in API Orchestration. We learned how to "chain" LLM outputs—using the result of one Gemini analysis to trigger a specific web search, then feeding that back into the model. We also gained a deep appreciation for the legal and technical nuances of Proxy management and the vital importance of data privacy for professionals working in high-risk environments.
What's next for PresstigatorThe future of Presstigator lies in Visual Intelligence. We plan to implement a dynamic "Connection Graph" using $O(n^2)$ relationship mapping to visually link entities on a map or timeline. We also want to expand our OSINT connectors to include satellite imagery analysis and leaked database archives (like the Panama Papers), making it the ultimate all-in-one workstation for the global press.
Built With
- fastapi
- nextjs
- python
Log in or sign up for Devpost to join the conversation.