M-Trix
User: Young professionals and students in Azerbaijan (22-35) who pay for rent, utilities, phone, loans, gym and several subscriptions (Netflix, Spotify, YouTube Premium) from one or two bank cards. Problem today: their payments are scattered across bank SMS messages, receipts and memory. Nobody tracks them in one place, so they miss deadlines (insurance, utilities, loan installments), get hit by "heavy weeks" when rent, a loan installment and other bills land within days of each other, and keep paying for duplicate subscriptions (e.g. Spotify + YouTube Premium) for months without noticing. Cancelling a subscription is also confusing: people don't know the steps or how to write the request. Existing banking apps show past spending, but they don't warn about what is coming or help cancel. Outcome: PayPulse turns an SMS, a receipt photo or one line of text into a tracked payment, then warns about upcoming deadlines, heavy weeks and duplicate subscriptions, and helps cancel unwanted ones.
Preliminary score
- Value for the user17.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI18.0 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing11.0 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility6.5 / 15
Data requirements, running costs and a clear next step.
- Originality5.0 / 10
A meaningful difference from common solutions.
What the judges wrote
- GPT-6 AstraPreliminary63/100
- Strengths
- Your proposed workflow connects SMS/text intake with a payment calendar, heavy-week warnings and cancellation guidance for a clearly defined audience. Your AZ/EN/RU support makes the experience locally relevant. Your testing report provides concrete failure cases—including API errors and misleading receipt fallback output—rather than unsupported reliability claims.
- To improve
- Replace the fabricated receipt fallback with an explicit extraction failure and manual-entry option; a warning does not make invented payment details safe. Benchmark a labeled set of SMS messages and receipts, reporting extraction accuracy, outage behavior and comparison with manual entry. For the enterprise track, identify a company-owned workflow and run a small pilot measuring time saved or missed-payment reduction. Add per-user Gemini cost estimates and a realistic storage and privacy plan, including which payment data reaches the model.
- Claude Opus 5.5Preliminary52/100
- Strengths
- You name a specific user (22–35-year-olds in Azerbaijan paying many bills from one or two cards) and concrete pains: missed deadlines, 'heavy weeks' and duplicate subscriptions. Your prototype appears to cover a real flow: paste a bank SMS or upload a receipt, Gemini parses it, and the radar, calendar and weekly summary update, with AZ/EN/RU support. Your failure write-up is unusually honest. You state that the backup mode shows a placeholder receipt result that does not come from the photo, which is exactly the kind of risk judges want surfaced.
- To improve
- Measure parsing quality. Run a set of real Azerbaijani bank SMS formats (Kapital, ABB, Leobank and others) and receipts through Gemini and through your rule-based fallback, then report accuracy for each. Compare the results with how users track bills today. The fallback should never display a fabricated receipt result; show an empty form that asks for manual entry instead. Add per-user Gemini costs, a plan beyond localStorage, and a concrete next step, such as a bank partnership or a pilot with 20 users. The deck's statistics ($252 a year, 70% trial trap) need sources and local figures, and you should tie the product to the enterprise track, for example as a white-label bank feature with a retention metric.
The team
3.Full name: Akbar Jalilov
designer
Turan GojayevLead
developer
2.Full name: Mahammad Talibli
product