<dev />

Osz Gyorgy Csongor

Software Engineer focused on Applied AI

I learn by building. Over time, my work has moved from full-stack applications toward language models, agentic systems, and practical AI tools.

GitHub

About

After completing my Computer Engineering degree at the University of Oradea, I continued learning through personal software projects. Building things has always been the most natural way for me to understand how technology works and turn ideas into something useful.

Over time, my interest shifted from traditional full-stack development toward artificial intelligence. Recently, I’ve been exploring language models, agentic workflows, RAG systems, and the challenges of turning AI experiments into practical applications.

I’m now starting a master’s degree in Artificial Intelligence and Vision at the Technical University of Cluj-Napoca. I see it as an opportunity to strengthen the theory behind what I’ve been building and explore machine learning and computer vision more deeply.

Outside university, I’m usually experimenting with a new tool, improving one of my projects, or learning something by building a smaller version of it myself.

Projects

Real problems, real solutions. Every project here was built to learn something new or scratch a personal itch.

Featured Project

Tutor Chatbot

Problem

Students learning CS concepts often get answers handed to them directly, which prevents real understanding and retention.

Solution

Built a Socratic AI tutor using a LangGraph state machine with 4-level hint escalation, misconception tracking, sandboxed code execution via Judge0, and RAG-based PDF upload for personalized tutoring.

Outcome

Deployed full-stack AI application with streaming responses, multi-provider LLM support (Groq, Gemini), and a complete pedagogical feedback loop — live at tutor.csongorosz.com.

Python
FastAPI
LangGraph
LangChain
ChromaDB
Next.js
Judge0 API
Railway
Vercel

CodeReviewAgent

Shipped

Agentic code review pipeline using LangGraph state machines and LangChain, enabling multi-step LLM reasoning to identify issues and generate structured reports.

PythonFastAPILangGraphLangChainNext.js

Social Network

Shipped

Built a full-stack social media platform supporting user authentication, posts, follows, and personalized feeds.

PythonDjangoJavaScriptHTML/CSSSQLite

Stock Trading Web App

Shipped

Developed a finance web application that allows users to simulate stock trading with real-time price data and portfolio tracking.

PythonFlaskSQLiteHTML/CSSJavaScript

Auctions

Shipped

Created an e-commerce style auction platform with listing creation, bidding logic, comments, and watchlists.

PythonDjangoSQLiteHTML/CSSJavaScript

Bomberman (C++)

Shipped

Developed a 2D Bomberman-style game implementing game loops, collision detection, grid logic, and object rendering.

C++Game LogicOOP

Library Management System

Shipped

Desktop application for managing books, users, and borrowing workflows using object-oriented design principles.

C#.NETSQL

Mail

Shipped

Single-page web email client implementing asynchronous requests, dynamic UI updates, and mailbox state management.

PythonDjangoJavaScriptHTML/CSS

Infrastructure Cost Estimator

Shipped

Built a configurable cost estimation tool for infrastructure projects, implementing pricing rules and calculation logic to model real-world budgeting scenarios.

PythonBusiness Logic ModelingCLI App

Tech Stack

Tools and technologies I use to build full-stack web apps, AI tools, and production-ready software.

Core Stack

Next.jsTypeScriptReactTailwind CSSshadcn/uiVercel

Backend & AI

PythonFastAPILangChainLangGraphChromaDBHuggingFace EmbeddingsRAG PipelinesGroqGeminiOllamaDjangoFlaskC#/.NETC++Java

Databases

PostgreSQLSupabaseSQLiteSQL

Cloud & DevOps

RailwayDockerLinuxGitGitHub ActionsCI/CD

Engineering Concepts

REST APIsState MachinesRAG ArchitectureSSE StreamingTool-Use / MCPOOP

Contact

Let's build something together.

Have an idea, a project, or just want to connect? I'm always open to conversation.