Our project addresses the problem of disorganized expense data caused by the abundance of finance apps, which often provide poor insights due to manual and limited filter options. To solve this, we developed a robust financial management solution that utilizes advanced AI and automation. The solution begins with PDF bank statement parsing, where users upload statements through an HTML interface. The Python Flask server then sends these documents to the Upstage API for document parsing. The parsed data is retrieved, organized, and categorized using Solar LLM before being stored in an SQLite database. This ensures structured and meaningful financial data for better insights. Users can also ask customized financial questions via HTML forms, which are processed using TogetherAI’s Llama 3.1 70B Instruct Turbo model. The system generates a corresponding SQLite query to extract the required data, performs necessary calculations, and returns the relevant information. Future improvements include AI fine-tuning for more accurate results, enhanced data visualization tools, advanced query options for more detailed insights for real-time updates and seamless user experience. Integration with Hedera and direct banking APIs for tampered-proof data for legal and audit purposes.
EasyEmail AI is an innovative tool designed to automate and simplify the process of email drafting. Powered by IBM Watson's advanced AI models, the application generates contextually accurate and professionally worded emails based on user inputs like recipient name, subject, and key points. Whether for business correspondence, client communication, or internal updates, EasyEmail AI ensures that every email is clear, concise, and personalized. The project aims to save users time by automating repetitive tasks, allowing them to focus on more strategic work. With a user-friendly interface, EasyEmail AI turns email drafting into a hassle-free experience.
BlindNav is a navigation aid tailored for visually impaired individuals, combining state-of-the-art AI models for seamless interaction. By leveraging YOLOv10 for object detection and tracking, the app identifies both static and dynamic objects in real-time. Whisper captures voice input from users, while O1 generates context-aware, concise responses to user questions about their surroundings. Users can upload or capture images, and the app processes them to give auditory feedback based on object tracking results. BlindNav aims to provide an intuitive and practical navigation solution for visually impaired users, improving their independence and situational awareness.