i build, i ship, people use it every day.

From idea to the App Store: I build software real users rely on every day. Not demos — systems running in the field.

Mobile & Full-Stack Developer · Flutter · Node.js · Applied ML/XAI

See the projects Download CVsoon

2: products live in production
~120: daily active users
1: peer-reviewed paper

Projects

live in production

Two products I built end to end, both in the daily workflow of real users today.

Live on the App Store

PolyMind

An AI-powered English learning app.

A cross-platform Flutter app built on Riverpod state management, backed by a scalable Fastify + PostgreSQL REST API and an LLM hosted on Groq for low-latency, adaptive lesson generation.

What's inside

Riverpod state management: one codebase across iOS and Android
Scalable REST API on Fastify + PostgreSQL
Low-latency lesson generation via an LLM hosted on Groq
Hybrid content: AI generation plus an admin-approved lesson pool
Adaptive difficulty: exercise flow that follows the learner's level

Flutter (Riverpod)
Fastify (Node.js)
PostgreSQL
Groq LLM

View on the App StoreSource codesoon

UI Flutter · Riverpod
API Fastify · Node.js
Data PostgreSQL · Groq LLM

~120 residents · in daily active use

Dormitory Management System

A desktop app running an entire dormitory's operations.

A desktop application running every operation of a 120-bed dormitory. It removed manual record keeping entirely.

What's inside

Student registration and room allocation from a single screen
Daily attendance tracking with full history
Payment tracking: balances, collections and overdue status
Automated monthly financial reporting
One source of truth in PostgreSQL — no paper ledgers

C#
.NET
PostgreSQL

Source code

Dormitory Management System

UI C# · .NET (Desktop)
API Business logic layer
Data PostgreSQL

Research

peer-reviewed paper

Are XAI Explanations Reliable? A Stability Analysis of SHAP and LIME under Data Perturbations in Credit Risk Prediction

25th International Istanbul Scientific Research Congress, 2026
Oral presentation · published in the proceedings
With Dr. Ömer Yılmaz

We measured how stable SHAP and LIME explanations remain under Gaussian noise on an XGBoost credit-risk classifier. Robustness was evaluated with Top-K Jaccard, Spearman/Kendall correlation and local explanation consistency: SHAP preserved the most important features, while both methods degraded on full rankings.

DOI

0.95: ROC-AUC (XGBoost)
SHAP + LIME: methods compared
Top-K Jaccard: stability metric

Experience

two terms at Rootcodes

I worked on the product team as a Mobile Application Developer Intern; the work fell into three tracks.

Mobile Application Developer Intern

Rootcodes Tokat, Türkiye

06/2025 – 07/2025 · 01/2026 – 02/2026

01 / backend

Built the PVC quotation and order management backend end to end (TypeScript/Node.js + PostgreSQL): role-based access control and a configurable dynamic pricing engine.
02 / r&d

Prepared the technical proposal, feasibility reports and a working prototype for a TÜBİTAK 1507 grant application; the R&D project idea was adopted by the company.
03 / mobile

Shipped cross-platform Flutter features and improved the state management architecture of existing modules.

Skills

the end-to-end stack

The tools I reach for daily, from the interface down to the database.

Languages

Dart · TypeScript · JavaScript · Python · C#
Mobile

Flutter · Riverpod
Backend

Node.js (Fastify) · .NET · REST API design
ML / Data

XGBoost · SHAP · LIME · scikit-learn · pandas
Databases

PostgreSQL · MySQL · Firebase
Tooling

Git · GitHub · Docker · Postman · Vercel

Spoken

Uzbek (native)
Turkish (advanced)
Persian (advanced)
English (upper-intermediate)

Contact

get in touch

A new team, a freelance project or a research collaboration — the easiest route is below. I usually reply within a few days.

Tokat, Türkiye