Nazlıcan Şahin
Software engineer and computer science senior at Eskişehir Technical University. I build production software across .NET, iOS, Go and React Native, and I research how to make AI tools cost less to run.
About
I am a final-year computer science student at Eskişehir Technical University in Turkey, drawn to the parts of engineering where backend systems, mobile apps and applied research meet. My work spans .NET services, native iOS in Swift and SwiftUI, cross-platform React Native, and Go APIs designed with clean and hexagonal architecture.
I spent an Erasmus exchange semester at FHV Vorarlberg in Austria, and interned as a software engineer at OBSS and Doğuş Teknoloji, where I shipped .NET and iOS work alongside production teams. Outside of code I lead the local Google Developer Groups and Women Techmakers chapters, running events that bring more people into engineering.
Right now my focus is my thesis on the Model Context Protocol: cutting the token usage and cost of AI tool calls through selective context, prototyped in Python.
Experience
Software Engineering Intern
Worked inside a product engineering team, contributing to backend services and shipping features through review, testing and production deploys.
- Backend
- .NET
- Teamwork
Software Engineering Intern
Built .NET services and native iOS features, moving between server-side logic and Swift / SwiftUI client work on real products.
- .NET
- iOS
- Swift
Erasmus Exchange
An exchange semester at the University of Applied Sciences Vorarlberg, studying software engineering in an international cohort.
- Exchange
- Software Engineering
Lead, GDG & Women Techmakers
Lead the local Google Developer Groups and Women Techmakers chapters: organising talks, workshops and community events for student and early-career developers.
- Community
- Leadership
- Events
Skills
Projects
EsDev - GDG Eskişehir Platform
Community events platform for GDG Eskişehir. A shared Go backend (DDD, clean and hexagonal architecture, GraphQL + REST over PostgreSQL) powering a Flutter attendee app with QR check-in and a Next.js organiser console.
GoFlutterGraphQLPostgreSQLNext.jsPiton Restaurant Recommender
Location-aware restaurant and café recommender: detects the user's position, maps nearby places via OpenStreetMap, and explains its picks. Next.js 14 with Firebase Auth and Neon Postgres for persistence.
Next.jsTypeScriptPostgreSQLFirebaselive demopopcorndb
Native iOS app in Swift that browses movie data from The Movie Database API, structured with a clean networking and core layer.
SwiftiOSREST- 1stars
ESTÜ Announcement Bot
Python bot that scrapes Eskişehir Technical University announcement pages, deduplicates new posts, and pushes them to Telegram on a daily schedule.
PythonAutomationTelegram - 1stars
KudosBoard Smart Contract
On-chain kudos board in Solidity, deployed to the Sepolia testnet with a Hardhat workflow and a lightweight web demo for sending and reading kudos.
SolidityTypeScriptHardhat anime-interval-timer
A Swift iOS interval timer app, built to practise SwiftUI state, timing and a polished native interface.
SwiftSwiftUIiOS
Research
Spending fewer tokens on the same answer
My senior thesis looks at how much of an MCP tool's response a model actually needs. Tool calls often return large payloads that inflate the context window, drive up token usage, and raise the cost of every request. I prototype selective context methods in Python that trim and reshape those responses, keeping the signal the model uses and dropping the rest.
- GoalLower token usage
- LeverSelective context
- PrototypePython