Skip to content
  • Python83.0%
  • HTML7.9%
  • TypeScript6.7%
  • 2.4%
Shares computed from the real volume of code in the public repositories.

Trabzon, Türkiye

Metehan Ulusoy

Computer Engineering student. I design it, I build it, I ship it.

Currently interning

Projects

  • Tarım Konseptleri — screenshot of the live site
    01TypeScript

    Tarım Konseptleri

    Three directions, one foundation — a monorepo that ships three agriculture sites, each with its own identity.

    • Next.js
    • Tailwind
    • GSAP
    • Lenis
    • Three.js
    Visit site
  • CENG Arşiv — screenshot of the live site
    02TypeScript

    CENG Arşiv

    Course notes and past exam papers for a whole department, filed by year and by course instead of scattered across group chats.

    • Next.js
    • Supabase
    • Tailwind
    • TypeScript
    Visit site
  • Streamly — screenshot of the live site
    03TypeScript

    Streamly

    A streaming interface rebuilt from scratch — rails, rankings, episode feeds, and a catalogue that is entirely fictional.

    • Next.js
    • React
    • Tailwind
    Visit site
  • 04Python

    LLM Cost Autopilot

    A gateway that routes every request to the right model — same quality, 50%+ cheaper.

    • FastAPI
    • scikit-learn
    • OpenAI
    • Anthropic
    • Ollama
    View source
  • 05Python

    Hybrid-Search RAG Pipeline

    Dense + BM25 dual index, RRF fusion, cross-encoder rerank — and a gate that knows when to say I don't know.

    • FastAPI
    • BM25
    • Embeddings
    • Cross-Encoder
    • Docker
    View source
  • 06Python

    Failure Forensics

    Observability that tells you which step of a multi-step AI pipeline actually broke.

    • OpenTelemetry
    • FastAPI
    • Streamlit
    • Docker
    View source
  • 07Python

    Semantic Caching Proxy

    A drop-in OpenAI-compatible cache that recognises the same question asked differently.

    • FastAPI
    • Redis
    • Embeddings
    • Prometheus
    • Docker
    View source
  • 08HTML

    n8n Automation Library

    2,055 production-ready n8n templates you can import and run immediately.

    • n8n
    • LangChain
    • OpenAI
    View source
  • 09Python

    Credit Risk Scoring

    End-to-end default prediction with the metrics banks actually use — KS, Gini, PSI.

    • XGBoost
    • SHAP
    • FastAPI
    • Docker
    View source
About

Twenty projects in three months

Study
Computer Engineering, third year, RTEU
What I work on
Web, machine learning, backend services, automation
Status
Doing an internship
Languages
Turkish, English

I study computer engineering at Recep Tayyip Erdoğan University, in my third year. I started with automation, went on through machine learning, and now write both web products and backend services.

I am not tied to one domain. What interests me is the problem itself: risk scoring at a bank, the data model behind a social platform, or a single-file command-line tool.

In every repo I write I look for the same three things: type safety, tests, and a README somebody can actually read.

  • 20

    open-source projects

  • 79

    commits

  • 9

    languages

  • 1.1 MB

    of source code

Data pulled on 2026-07-25

How I got here

Every row is a real repository on a real date.

  1. March 2026

    Automation and first models

    The fastest way to understand something is to take it apart until it breaks.

    6 projects

  2. April 2026

    Applied machine learning

    Banking problems. This is where I learned that accuracy on its own says nothing.

    4 projects

  3. May 2026

    Services

    Seven services in three weeks: gateway, cache, observability, CI gate.

    10 projects

Contact

LET'S
BUILD