Executive summary
A finance product built around a relationship, not an individual account.
Budżetówka is a local-first household finance system for couples and married partners who need one shared financial picture but do not want their most sensitive data to depend on a commercial cloud platform.
I led every discipline myself: user and market research, business analysis, requirements, information architecture, UX flows, UI and design system, technical architecture, implementation and quality assurance.
The project is also a practical demonstration of full AI product building. I used AI to compress execution across research synthesis, specification, prototyping, coding and verification—while keeping product judgment, privacy decisions and financial rules under human control.
01 · Problem and niche
Shared life. Fragmented money.
Most personal-finance tools assume a single owner, a bank-connected cloud account, or complete financial separation. Couples often live in the gap between those models: two people, individual and shared accounts, joint obligations, different spending patterns and one household plan.
The available workarounds create their own cost. Spreadsheets require continuous maintenance. Bank aggregators ask for sensitive access. Generic budgeting applications rarely explain who paid, what is shared and how a decision affects both people.
02 · Product principles
Privacy is not a setting. It is the architecture.
Local source of truth
Primary financial data lives in a local SQLite database. Core workflows remain available without a network connection or cloud account.
Explicit sharing
Nothing leaves the device by default. Optional synchronization, backup or AI support must be consciously enabled and scoped by the user.
Secrets stay separate
Credentials and tokens belong in the operating system’s secure credential storage—never in the finance database or source repository.
Deterministic money
Amounts are stored as integer grosze. Financial calculations run in testable code; language models never calculate or silently write money.
Human confirmation
Imported or AI-extracted records pass through review before persistence. The product shows the decision and keeps correction possible.
Portable by design
The user can understand and move their data. Cloud services are optional extensions, not a condition for accessing the household’s history.
Trusted boundary · the user’s device
Optional extensions
Encrypted backupExplicit syncOpt-in AIDiagram placeholder · replace with final security and data-flow visual
03 · My role
One product. Every discipline.
The point was not to produce a concept deck. It was to move one real product through every layer required to make it coherent, testable and usable.
- Discover
User and market research
Interviewed prospective users, examined household workflows and mapped the privacy compromises in existing alternatives.
- Define
Business analysis and requirements
Translated needs into functional requirements, non-functional constraints, domain rules, risks and a phased roadmap.
- Structure
Information architecture and UX
Designed onboarding, zero-based budgeting, ledger, import, recurring payments, couples comparison and settings flows.
- Systemise
UI and design system
Created tokens, reusable components, responsive screens, states and documentation across 26 Figma pages in Polish and English.
- Build
AI-assisted implementation
Used vibe coding and agentic workflows to turn specifications into working Swift and cross-platform Tauri code.
- Prove
Testing and release
Verified financial invariants, database migrations, product builds and the macOS installer instead of treating prototype fidelity as completion.
04 · Product process
AI accelerated the work. It did not own the decisions.
I treated AI as a production layer spanning research synthesis, specifications, interface exploration, code generation, test creation and documentation. Each task had bounded context, a clear output and a verification step.
The boundary was deliberate: AI could propose and execute; I remained accountable for the product model, ethical choices, security posture, usability and what counted as done.
Vision · user needs · privacy · trade-offs · acceptance
Synthesis · variants · scaffolding · coding · test coverage
Builds · schemas · calculations · links · regressions
05 · Product experience
Designed for two financial realities and one household plan.
Zero-based budget
Give every grosz a job, carry selected categories forward and keep the household plan explicit.
Individual and shared views
Compare each person, joint accounts and the whole household without losing context.
Ledger and import
Add, review, categorise and reconcile transactions with clear ownership and duplicate detection.
Recurring payments
Turn repeating costs into a forward-looking plan rather than another notification.
Cash flow and reports
Explain where money changes, why it changes and how it affects both partners.
Lock and private settings
Keep local access, credentials, sharing and automation under explicit user control.
06 · Evidence and current state
From research artefacts to a working system.
- Shipped foundation
Native macOS product
The proven v2 application covered the complete household-finance workflow, passed 133 core tests and was packaged as a verified DMG installer.
- Design source
Complete bilingual case file
The Figma source contains system context, information architecture, user flows, ERD, feature catalogue, 11 application screens, 9 modal states and user journeys.
- Active evolution
Cross-platform v4
The current Tauri, React, TypeScript and Rust rewrite has a verified application scaffold and SQLite foundation. The financial core is being ported with tests.
The product is the proof: I can move from an ambiguous human problem to requirements, interaction, system architecture and working software—without outsourcing coherence between the disciplines.
07 · What this demonstrates
Product leadership at builder depth.
For recruiters
End-to-end product ownership, design leadership, systems thinking, privacy by design and the ability to collaborate credibly with engineering.
For business partners
A single partner who can reduce ambiguity, define the product, make it tangible and carry it into implementation.
For product teams
A practical AI-native workflow where faster execution increases evidence and quality instead of replacing judgment.