EFECAN KÜÇÜK

3 branches · 0 merge conflicts

efecan@istanbul ~ $ git branch -a

* data-analytics # full-time at Foneria — dashboards, anomaly detection, SQL mobile-development # two platforms in production — SwiftUI & Jetpack Compose ai-engineering # RAG, MCP, multi-agent — the branch pointed at the future

efecan@istanbul ~ $ python analyze.py --target=efecan

[ INFO ] loading 3 branches... OK

[ SCAN ] running modified z-score (MAD) on subject...

[ALERT ] anomaly detected: 3 careers in 1 person (z = 4.2)

[ OK ] false_positive: false — confirmed by production data

efecan@istanbul ~ $ ▊

Browse projects Download CV (PDF)

$ ./anomaly-engine --live --input=pointer

Live Anomaly Feed

scanning… all clear

your pointer movement is a live metric — my MAD-based engine flags the spikes anomalies caught: 0

$ ls projects/ --branch=data-analytics

Data Analytics

anomaly-detection-engine

Three-layer real-time monitoring for financial streams. Modified Z-Score over MAD with day-parting baselines — 60% fewer false positives.

React · NestJS · PostgreSQL · OpenRouter

pdf-pipeline

Extracted portfolio data from 489 fund disclosure PDFs with OCR fallback and LLM-based error correction. Replaced a monthly Excel ritual.

Python · pdfplumber · Pandas · PostgreSQL

madwatch OPEN SOURCE

The anomaly math from production, packaged — pip install madwatch. MAD-based Modified Z-Score, seasonal baselines, a streaming detector and a CLI.

Python · NumPy · pytest · GitHub Actions · PyPI

$ ls projects/ --branch=mobile-development

Mobile Development

Foneria

A production fintech investment app, twice — SwiftUI on iOS, Jetpack Compose on Android. Real money, real users, real code review.

Swift · SwiftUI · Kotlin · Jetpack Compose · Hilt · Firebase

crypto-portfolio-tracker OPEN SOURCE

A personal crypto portfolio app built end to end — SwiftUI on the phone, my own Go + Postgres API behind it, price alerts over APNs.

Swift · SwiftUI · Go · PostgreSQL · APNs

$ ls projects/ --branch=ai-engineering

AI Engineering

fund-discovery-api

RESTful fund discovery with 15+ filter criteria, feeding an AI chatbot. Multi-agent architecture for autonomous fund analysis on the way.

NestJS · TypeScript · PostgreSQL · Docker · Swagger

everything-claude-code OPEN SOURCE

An agent harness optimization system — skills, instincts, memory and research-first development patterns for Claude Code, Codex, Cursor and beyond.

Claude Code · Agent Skills · MCP · Markdown

$ cat skills.txt --bars

Skill Levels

SQL · Python 92%

Power BI · Looker 88%

RAG · MCP · n8n 85%

Airflow · ETL 82%

Kotlin · SwiftUI 78%

Escaping Excel 97%

$ cat about.md

About

Data analyst at Foneria by day. Two production mobile apps on the side. Lately teaching agents to do the boring parts. Three careers, zero merge conflicts — each one gets its own branch.

full history →

$ tail -3 blog/

Latest Posts

[data] Why MAD instead of standard deviation? May 2026

all posts →