FynX | Devpost

The Problem We Couldn't Ignore

Pakistan has 65 million people using digital banking. 5.2 million small businesses contributing 40% of the country's GDP. And almost none of them have real visibility into their own money.

We watched people around us freelancers, shopkeepers, small factory owners juggling paper registers and four different apps, abandoning every finance tool within weeks because typing every transaction by hand just doesn't stick. Banks reject 70–80% of SME loan applications not because the businesses aren't viable, but because they can't prove their cash flow. There's no record. No history. Nothing.

That's what FynX started from.

What We Built

FynX is two purpose-built AI applications sharing one Gemini intelligence layer.

FynX Personal reads incoming bank SMS alerts in the background and logs every transaction automatically. No typing. The Gemini-powered Co-Pilot generates a real-time Financial Health Score from 0 to 100, forecasts month-end spending, detects forgotten subscriptions, and answers financial questions in plain language from actual user data.

FynX Business gives Pakistan's 5.2 million MSMEs the financial infrastructure they never had. Automated accounting, inventory, HR, payroll, and CreditIQ an alternative credit scoring engine that builds a credit profile entirely from real transaction behavior:

$$\text{CreditIQ} = 0.35 \cdot \text{CF}{\text{consistency}} + 0.25 \cdot \text{SR}{\text{rate}} + 0.25 \cdot \text{VPR} + 0.15 \cdot \text{Activity}$$

The Business CFO Chat lets owners ask "Can I afford to hire someone next month?" and get answers from their actual numbers via Gemini function calling.

How We Built It

Flutter for Android and iOS. Firebase for backend and real-time sync. Google Gemini API with a multi-model fallback chain across three variants with exponential backoff for 99% AI availability. Google ML Kit for on-device OCR receipt scanning. Android BroadcastReceiver for background SMS parsing across HBL, Meezan, UBL, MCB, Bank Alfalah, and EasyPaisa formats. Hive for offline caching. Clean Architecture with BLoC state management.

Challenges We Faced

Pakistani bank SMS formats are inconsistent three different formats from Bank Alfalah alone. We built a pattern library with manual correction fallback to handle every edge case. Voice tracking accuracy dropped on Urdu-English mixed input. We fixed this by prompting Gemini to explicitly handle code-switched language, which is how most Pakistanis actually speak. CreditIQ prompt engineering took around 30 iterations before Gemini produced reliable, explainable scores in a strict JSON schema rather than unstructured prose.

What We Learned

Alternative credit scoring from behavioral transaction data works. Early testers with zero formal credit history got CreditIQ scores that matched our manual assessment of their financial health. For a country where most small business owners are invisible to formal credit bureaus, that result matters more than the app itself. Gemini function calling changed how we think about the product entirely letting users talk to their finances rather than navigate menus makes FynX feel like an advisor, not a ledger.