Full-stack & AI software developer
Farid Mahmudlu
I build software products end-to-end — from the data model and auth to the interface, the AI pipeline and the production deploy.
Scroll — follow the signalOrigin
One person, the whole request path.
I study Computer Science at Eötvös Loránd University in Budapest. In 2026 I founded or co-founded five software products, and in each of them I owned the parts most teams split across several people: schema, auth, services, interface, AI integrations and the deploy.
Every request a user makes travels through all of those layers. Follow the signal down this page — it passes through each one.
- 5 products founded or co-founded in 2026
- 6 layers owned end-to-end
- 8 programming languages
- 5 spoken languages
L01 Data
Schemas before screens.
Most product bugs are data bugs that surface late. I start with the model — constraints, transactions and migrations that make invalid states impossible to store.
- StockFlow
Inventory mutations run in isolated PostgreSQL transactions, and every stock change writes its audit-log entry inside the same transaction.
- Calisiyo
Supabase schema with migrations, triggers and Realtime behind study plans, goals, mock-exam tracking and spaced review.
- Parabola
JPA / PostgreSQL domain model for users, body profiles and garments that feeds the size-recommendation engine.
L01 / Data
- Engines
- PostgreSQL · Supabase · Firestore
- Access
- Prisma · JPA
- Patterns
- Atomic transactions · audit logs · migrations · triggers
L02 Auth & Security
Trust is a layer, not a feature.
Access rules belong where the data lives, not only in the UI. I enforce them in the database and the API first, then make the client reflect them.
- Hushplan
Private spending limits live in a Postgres schema the API never exposes, reached only through membership-checked functions — every privacy rule covered by pgTAP tests.
- ShareVibe
Firestore and Storage security rules generated from a single access list, with separate owner and admin roles per cafe workspace.
- StockFlow
JWT / Passport authentication, bcrypt hashing and three-role RBAC — admin, manager, staff — across API and UI.
- CanvasFlow
Third-party API tokens encrypted at rest with AES-256-GCM and never sent to the browser.
L02 / Auth & Security
- Identity
- OAuth 2.0 / PKCE · JWT · Clerk · Auth.js · Firebase Auth
- Authorization
- RLS · RBAC · security rules · storage policies · Spring Security
- Hardening
- AES-256-GCM · bcrypt · rate limiting · security headers
L03 Backend
Boring services, on purpose.
Predictable APIs with explicit contracts, real-time where it actually matters, and business logic that is easy to test and easy to reason about.
- Parabola
Java 17 / Spring Boot services documented with OpenAPI, including a fit engine that scores body-profile compatibility and recommends a garment size.
- StockFlow
NestJS backend that broadcasts stock activity to every connected client in real time over Socket.IO.
- CanvasFlow
Sync engine that reads Canvas LMS rate-limit headers and adapts its polling interval to stay under quota, with a time-budget guard for serverless limits.
L03 / Backend
- Runtimes
- Node.js · Java 17 · Python
- Frameworks
- NestJS · Express · Spring Boot
- Interfaces
- REST · OpenAPI · Socket.IO · webhooks
L04 Interface
Where the system meets a person.
The interface is the only layer users ever see. I build it in React, Next.js and React Native — fast, accessible, and honest about loading and error states.
- Calisiyo
Next.js flows for study planning, Pomodoro, goals, resources, mock-exam tracking and spaced review.
- ShareVibe
QR-based, table-specific guest flows where guests upload and react to media and unlock campaign rewards.
- StockFlow
Expo / React Native app with swipe navigation and memoised lists built for 1,000+ items.
- Hushplan
Flutter app in seven languages, tested on six screen sizes at two text scales so no screen ever overflows.
L04 / Interface
- Web
- React · Next.js · Vite · Tailwind CSS
- Mobile
- Flutter · React Native · Expo
- State
- Redux Toolkit · Zustand
L05 Intelligence
Models are components, not magic.
I treat AI models like any other dependency: typed inputs, structured outputs, swappable providers — and a pipeline around them that I can actually debug.
- Avalabs
Multimodal pipeline: FFmpeg extracts frames and audio, Whisper transcribes, and GPT, Gemini or Claude analyse the result with Instagram context.
- Hand Tracking
Low-latency MediaPipe + OpenCV hand tracking with One Euro filtering, driving gesture-based OpenGL particles — the ancestor of the particles on this page.
- QuickScript AI
Structured short-form video scripts in six languages, with tone and platform controls.
L05 / Intelligence
- Models
- OpenAI · Whisper · Gemini · Claude · OpenRouter
- Media & vision
- FFmpeg · MediaPipe · OpenCV
- Patterns
- Structured outputs · provider switching · multimodal pipelines
L06 Production
Shipped is a feature.
Code that is not deployed does not exist yet. I own the last mile: hosting, HTTPS, caching, monitoring and the tests that guard every release.
- ShareVibe
Production deployment on a VDS behind Nginx, with HTTPS, caching and security headers.
- Calisiyo
PostHog analytics, Sentry error monitoring, web push and Playwright end-to-end validation.
- Parabola
Supported the public launch and the team’s first clothing-store onboarding.
- Hushplan
Version 1.0 in Google Play internal testing, CI on every push and daily encrypted database backups.
L06 / Production
- Hosting
- Vercel · VDS / Nginx · Docker
- Observability
- Sentry · PostHog
- Quality
- Playwright · unit & integration tests
Shipped
Products & projects
Five products founded or co-founded in 2026 — each built end-to-end — and the projects around them.
Founded & co-founded products
- Hushplan Group plans that fit everyone’s budget, with private spending limits. Case study:
- Calisiyo Study-planning SaaS for Turkey’s YKS university entrance exam. Case study:
- Parabola Recommendation-based fashion marketplace with a size-fit engine. Case study:
- ShareVibe Multi-tenant, QR-based social platform for cafes. Case study:
- Avalabs AI social-media intelligence for Instagram content. Case study:
Selected projects
- StockFlow Inventory system with atomic transactions and a mobile app. Case study:
- CanvasFlow Encrypted, near-real-time Canvas LMS sync for students. Case study:
- QuickScript AI AI script generator for TikTok, Reels and Shorts. Case study:
- Real-Time Hand Tracking Real-time hand tracking driving GPU particle effects. Case study:
More experiments
- RepRecord (opens in a new tab)
Telegram bot for workout logging and progress charts.
Python · Telegram Bot API · PostgreSQL · Matplotlib
- Mockwise (opens in a new tab)
AI-generated product mockups.
TypeScript · AI
- TriCharge (opens in a new tab)
E-commerce platform with Stripe payments and JWT auth.
React · Express · Supabase · Stripe · JWT
- Cookly (opens in a new tab)
Personalised recipe generation with AI.
TypeScript · AI
- Last Flame (opens in a new tab)
Cooperative escape game for 2–4 players, built at a hackathon.
TypeScript · Multiplayer
Toolkit
The stack, all of it.
Languages
- TypeScript
- JavaScript
- Python
- Java
- Dart
- C
- C++
- C#
Frontend / Mobile
- React
- Next.js
- React Native
- Expo
- Flutter
- Vite
- Tailwind CSS
- Redux Toolkit
Backend / Data
- Node.js
- Express
- NestJS
- Spring Boot
- PostgreSQL
- Supabase
- Prisma
- Firebase / Firestore
- JPA
- pgTAP
AI / Systems
- OpenAI
- Whisper
- Gemini
- Claude
- OpenRouter
- FFmpeg
- MediaPipe
- OpenCV
- Socket.IO
- JWT / OAuth
- Docker
- Playwright
- Sentry
- PostHog
Log
Timeline
-
Released Hushplan 1.0 to Google Play internal testing.
-
Founded Hushplan, a privacy-first group-planning app. Started CanvasFlow.
-
Co-founded Calisiyo, a study-planning SaaS for the YKS exam.
-
Co-founded Parabola. Started StockFlow. Earned Google AI Essentials.
-
Co-founded ShareVibe, a QR-based social platform for cafes.
-
Co-founded Avalabs and led its entire technical implementation.
-
Built QuickScript AI, a multilingual short-form script generator.
-
Built real-time hand tracking with MediaPipe, OpenCV and OpenGL.
-
Started a BSc in Computer Science at ELTE, Budapest.
FAQ
Quick answers
Who is Farid Mahmudlu?
Farid Mahmudlu is a full-stack and AI software developer based in Budapest, Hungary, and a Computer Science BSc student at Eötvös Loránd University (ELTE), expected to graduate in 2028. In 2026 he founded Hushplan and co-founded four software products — Avalabs, ShareVibe, Parabola and Calisiyo — and built each of them end-to-end.
What does Farid build?
Web and mobile products from the database up: data models and authentication, backend services, React, Next.js, React Native and Flutter interfaces, AI pipelines using models such as GPT, Gemini, Claude and Whisper, and production deployment with monitoring.
Which technologies does he work with?
Mainly TypeScript, React, Next.js, React Native, Flutter, Node.js, NestJS, Spring Boot, PostgreSQL, Supabase, Prisma and Firebase. For AI and media: OpenAI, Whisper, Gemini, Claude, OpenRouter, FFmpeg, MediaPipe and OpenCV.
Which products has he founded?
He founded Hushplan (privacy-first group planning, 2026 — ongoing) and co-founded Avalabs (AI social-media intelligence, 2026), ShareVibe (QR-based social platform for cafes, 2026), Parabola (recommendation-based fashion marketplace, 2026) and Calisiyo (study-planning SaaS for the YKS exam, 2026 — ongoing).
Where is he based, and which languages does he speak?
He is based in Budapest, Hungary. He speaks Azerbaijani (native), Turkish (C1), English (B2), Russian (A1) and Hungarian (A1).
How can I get in touch?
By email — the address is in the contact section of this page — or via LinkedIn. His public code is on GitHub.
Contact
Open a channel.
Roles, collaborations, or a product that needs one person to own it end-to-end — email is the fastest way to reach me.