Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{194504,
        author = {Tanishq Vishwas Tule and Atharva Shridhar Shirke and Aarya Shridhar Shirke and Sanchit Patil},
        title = {Flux: An AI-Powered Personal Finance Management System with On-Device Machine Learning and Conversational Agent Integration},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {10},
        pages = {5084-5092},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=194504},
        abstract = {Personal finance management remains a persistent challenge for individuals owing to cognitive biases, reactive decision making, and the absence of real time intelligent insights in conventional budgeting tools. This paper presents Flux, an Android personal finance application that bridges the gap between passive expense trackers and proactive financial advisors. Flux combines on device machine learning comprising a TensorFlow Lite transaction classifier (FluxBrain), a statistical anomaly detector using the two-sigma rule, a linear regression spending predictor, and a merchant interval recurring payment detector with a cloud hosted conversational AI agent powered by Google Gemini 2.0 Flash. The agent employs an agentic function calling loop with eleven domain specific tools spanning affordability assessment, budget health analysis, category trend analysis, cash flow forecasting, web augmented price discovery via the Tavily API, and natural language transaction ingestion. Persistent per user memory and multi session chat history are managed in Firebase Fire store. The Android client is implemented in Kotlin with Jetpack Compose following MVVM architecture and Hilt dependency injection. Security features include biometric authentication via Android Biometric Prompt, FLAG SECURE display protection, certificate pinning over Ok Http, and root/emulator integrity checks. Evaluation reveals that the hybrid on device/cloud architecture achieves sub second categorisation latency for routine transactions while delivering richer analytical depth through the generative AI backend. Flux demonstrates how integrating statistical machine learning, large language model function calling, and real time cloud data can produce a compelling intelligent personal finance assistant.},
        keywords = {Personal finance management, Android application development, machine learning, generative AI, TensorFlow Lite, Google Gemini, Firebase Fire store, conversational agent, anomaly detection, budget management, Jetpack Compose, MVVM architecture.},
        month = {March},
        }