Portfolio · Web, mobile and desktop
Pakistan
Jahanzeb
Mohsin
Full-stack developer
13 projects you can actually run.
- 1AI tools that cite their sources.
- 2Apps that work offline.
- 3A Windows application published on the Microsoft Store.
Case studies
Selected work
Seven projects written up in detail: what each one does, how it works and the decisions behind it. This is a selection; client work and some personal projects stay private.
Also on GitHub
Smaller apps and tools. Each one has a README with screenshots.
Started as client work, rebuilt as my own product
Invoicing for Windows, macOS, Linux, web and mobile, with offline storage, optional cloud sync, PDF invoices and Stripe payment links.
Flutter · SQLite · Supabase
Microsoft Store ↗
A rotating shift calendar that works out pay, including overtime and holiday rates, and checks it against your payslip.
Flutter · Drift · Riverpod
Quizzes for classes and study groups, with admin tools, CSV import, timed attempts, leaderboards and results analytics.
Flutter · Firebase
A desktop lead finder for a keyword and a city, with contact extraction, CSV and Excel export, and a native module written in Rust.
Electron · Playwright · Rust
A WordPress plugin that sends order updates to customers on WhatsApp in the background and tracks delivery through webhooks.
PHP · WordPress · React
An offline Android app for tracking subscriptions, what they cost each month and when the next payment is due.
Flutter · SQLite
Equipment and invariants
The kit, and the rules it follows
The tools I reach for, and rules the code itself enforces. Each rule points to the project that shows it.
Equipment
- Web
- TypeScriptNext.jsReactNode.jsTailwind CSSAstro
- Mobile and desktop
- FlutterDartRiverpodBlocElectronTauri
- Data and backend
- PostgreSQLpgvectorPrismaSupabaseSQLiteFirebaseFastAPI
- AI
- Agents and tool callingMCPRetrieval (RAG)EmbeddingsGuardrails and human approvalPrompt and model evaluationOpenAI API
- Also
- RustPythonPHP and WordPressDockerPlaywright
Invariants
- 1
Money is stored in the smallest currency unit, never as a floating point number.
Dukaan Book → - 2
Every AI answer shows its source or a confidence score.
ScopeSignal, Support Copilot → - 3
Every tool call is checked in code. Even a fooled model cannot pay an unverified account.
CarryPay → - 4
Below 0.6 confidence, a person makes the call.
ScopeSignal → - 5
Money moves only after the family approves with a passkey that signs the exact payment.
CarryPay → - 6
The AI drafts the reply. A person sends it.
AI Support Ops Copilot → - 7
Models are picked on a labelled test set, not by feel.
ScopeSignal → - 8
The encryption key never leaves the Rust side.
Notes Vault → - 9
No server. Computers find each other on the local network.
LAN Messenger →
About
How I work
I started out building Flutter apps and Python tools, including a few computer vision projects. Since then most of my work has moved to TypeScript and AI-powered web apps, and I still build mobile apps in Flutter.
When an app uses AI, I test the prompts and models against fixed examples before trusting them, and I keep a person in the loop wherever a wrong answer would cost someone money or trust. Several of the case studies above show how that works in practice.
Record
Education
- Master of Information Technology2022-2024
- Bachelor of Commerce (B.Com)
Certificates
- IBM Applied AI Professional CertificateIBM, through Coursera
- Flutter AppsEhunar
Contact
Write to me
- jahanzeb-mohsin
- GitHub
- JahanzaibMohsin9