SafeBalance
Your friendly cash companion
By Team Harkonnen

Background
The gig economy represents the sector of the economy with workers engaged in flexible, short-term jobs often involving day-to-day payments. Gig economy workers are often hired on a temporary basis for certain , and their wages vary widely depending upon the quantity of tasks they receive per day, and the payout for each task. This is unlike the full-time employment offered in traditional white or blue-collar jobs, which offer a stable contractually mandated monthly salary. Examples of blue collar work include delivery services, drivers for cab aggregators like Ola and Uber, freelance work like graphic design or writing, people engaged in part-time labour intensive work as farm labour and so on. Per a 2022 policy brief by NITI Aayog, India had approximately 7.7 million gig and platform workers in 2020-21, a figure projected to surge to 23.5 million by 2029-30.

Featureset

Income Volatility Prediction Engine
The heart of the platform will be an Income Volatility Prediction Engine, a collection of lightweight machine learning based predictive models built using historical data that will adapt itself to the user’s spending habits. The cash flow will be recorded from multiple sources; UPI apps (like GPay or PhonePe), bank accounts transaction histories and manual cash entries. By unifying fragmented financial data, the agent will build a profile of a user’s income and expenses. It will parse spending from transactions, classify it into categories (rent, EMIs, food, travel), and continuously update the user’s financial profile. Based on this data, the users will be classified into user types. The users will be classified on the basis of their income levels, their vulnerability, the variance in their monthly incomes, and their willingness to adjust their financial habits.

Conversational Chat Interface
The user will be able to interact with the AI agent through a chat interface. The agent will be multilingual and will support conversations in multiple Indian languages, with a voice mode for easier accessibility. Users can ask questions like,
How much can I spend this weekend?
Or
How much money should I save for my child’s birthday next month?,
without going through complex planning or navigating through spreadsheets of account data.

Insights
The agent will deliver short snippets of financial suggestions in the form of “Insights”. These are short notifications that suggest immediate action, delivered in-app or through email/SMS. The actions suggested will be intuitive and can be easily executed by the user. For example,
Phone EMI payment next week, move 800₹ to savings.
Or
Expected increase in expenditure at the end of month, conserve money by reducing spending.
These Insights will be capped to a certain amount (around 10–20 per month) in the free tier to prevent user annoyance, and the Insight limit could be increased in the paid tier.

Cashflow Forecasting and Budgeting
SafeBalance will predict the cash flow using time series modelling of the user’s transactions. Using this it can predict, with a certain level of certainty, the likelihood of the user being in a financially unsafe position at some point in the near future. For example,
69% chance of bank balance dropping below 1000₹ by October 15th
Each forecast will come with an explanation and suggested actions for the user. Based on whether the user carries out the action, we can also figure out how receptive the user is to suggestions and how likely they are to execute them.

Privacy and Data Control
SafeBalance will follow stringent privacy and security practices for handling personal data in accordance with the Digital Personal Data Protection Act, 2023 and the Sensitive Personal Information Rules, 2011. User data will be anonymised. Every data integration will be opt-in. Users can view the data accessible to the platform and delete it anytime they want.

Business Model and Revenue
The platform would utilize a freemium business model, with a free tier and paid tier with a monthly subscription focused on more enhanced Insights. An additional revenue stream could be through partnerships with banks and other financial institutions and services, enabling integrated services and potential collaboration of ideas. At a more advanced stage, the platform could also offer investment advice for the users to invest their spare cash into.

Tech Stack
Flutter app(cross-platform for iOS & Android)
Firebase Auth (Google, Email/Password, Phone OTP)
MongoDB Atlas (Transactional Data (Time-Series))
PostgreSQL (User data and analysis)
IVP Engine (LSTM(TensorFlow/PyTorch))
Flask and Docker for endpoints and containerization
Azure for deployment
UPI API for transaction access.