INVESTI

AIdriven platform designed to streamline the startup launch process through stepbystep guidance. It also analyzes financial documents to detect fraud or scams and automatically generates a PPTX report

Comment

Inspiration

The startup journey is often overwhelming for aspiring entrepreneurs, especially when navigating legal, financial, and strategic hurdles in tunisia. We were inspired to create INVESTI to simplify this process and empower innovators with AI-driven guidance. Additionally, we recognized the increasing risk of financial fraud and scams in early-stage investments—so we built a solution that not only guides founders but protects them as well.

What it does

INVESTI is an AI-powered platform that: -Guides users step by step through launching a startup. -Reviews financial documents such as balance sheets, investor agreements, and invoices. -Detects red flags related to fraud or scams using advanced NLP and anomaly detection. -Automatically generates a PowerPoint report summarizing findings, recommendations, and potential risks.

How we built it

We built INVESTI using a modular tech stack: Frontend: Django for a responsive and intuitive UI. Backend: Python AI Models: llama 3.2 for startup guidance, and fine-tuned NLP models for fraud detection. Document Processing: PyMuPDF and LangChain for parsing and analyzing PDFs. Report Generation: python-pptx to auto-generate PowerPoint summaries. Database: Faiss for storing user sessions and insights.

Challenges we ran into

Designing a generalizable fraud detection system that works across many document types was difficult due to data diversity. Ensuring accuracy in AI-generated business advice while keeping responses user-friendly required careful prompt engineering and testing. Integrating all components into a smooth user workflow was more complex than expected, especially report generation from mixed content.

Accomplishments that we're proud of

Successfully built an end-to-end pipeline that turns raw user documents into structured startup guidance and fraud analysis. Created a system that can automatically generate pitch-ready PowerPoint slides—a major time-saver for early-stage founders. Helped a few beta users identify discrepancies in investor documents during early testing.

What we learned

Startup founders need simplicity and clarity, not just features—so usability is everything. Combining LLMs with traditional rule-based fraud detection gives a more reliable and explainable outcome.

What's next for INVESTI

Expand the AI guidance engine to cover product-market fit, fundraising, and hiring strategies. Enhance fraud detection by training on real-world scam cases and expanding document formats (Word, Excel). Integrate with platforms like Stripe Atlas or Mercury to support startup formation directly. Launch a mobile app for on-the-go document review and alerts.

Built With

Share this project:

Updates