Spark
Introducing Spark your ultimate AIpowered financial Mentor! This innovative app seamlessly uses Omi chat interface, conversation analysis and external integration to provide financial advice/tracking.
Created on 1st April 2025
•
Spark
Introducing Spark your ultimate AIpowered financial Mentor! This innovative app seamlessly uses Omi chat interface, conversation analysis and external integration to provide financial advice/tracking.
The problem Spark solves
Spark is a hands-free, AI-powered financial mentor that simplifies expense tracking, budgeting, and crypto management by integrating with Omi, a wearable audio transcriber. Users can verbally log transactions—like saying, “I spent $300 on clothes”—and Spark and Omi instantly record the expense, provide real-time financial advice, and sync data to an external dashboard that visualizes spending trends. Inside the Omi app, Spark features an AI chat assistant that uses the user’s own expenses as context to offer personalized financial guidance. By soon integrating with MetaMask, Spark will also track crypto portfolios, offering a seamless Web3 financial overview. This eliminates the need for manual data entry, making financial management effortless, while AI-driven insights help users make smarter money decisions. Additionally, Spark enhances security by reducing reliance on banking apps and spreadsheets, minimizing exposure to phishing attacks and manual errors.
Challenges we ran into
One of the biggest challenges we faced while building Spark was deploying the backend as a public API. Initially, we built the backend using Hugging Face’s NLP libraries and transformers to process financial conversations. However, when we attempted to deploy this setup on a cloud server, we ran into major issues—the model’s requirements were too large, making it difficult to run efficiently on a typical AWS instance. The server struggled with memory limits, slow performance, and dependency management, which created a significant deployment roadblock.
To overcome this, we had to rethink our approach and develop a lighter solution. Instead of relying on Hugging Face’s heavy NLP models, we streamlined the backend by leveraging a simpler OpenAI prompt combined with financial data. This allowed us to maintain AI-driven insights and a conversational experience while significantly reducing the resource load.
Alongside this, we set up an Ubuntu instance on AWS, carefully optimized the environment, and configured security groups and networking to ensure smooth deployment. By utilizing AI tools, diving into documentation, and applying creative problem-solving, we successfully overcame these challenges and deployed Spark’s backend as a public API.
Tracks Applied (1)
Most Helpful AI Assistant / Mentor
Technologies used