ece.

BERLIN · DATA × PRODUCT × AI

From complex data
to something
that matters.

I’m Ece. I design reliable data systems, own the decisions behind them, and build products that bring them to life.

Explore my work ↘The person behind it →
4+ yearsin data engineering & governance
8+ yearsacross technology, quality & compliance
End to endfrom architecture to implementation

01 / SELECTED WORK

Three questions.
Three things I’m building.

Personal projects at the intersection of reliable systems, thoughtful products and applied AI.

3 projects · different stages, shared curiosity
01 / DATA ENGINEERING / APPLIED AIIn development

YouTube Opportunity Finder

What’s worth creating next?

Turning fragmented discovery signals into a structured way to explore Turkish YouTube content opportunities.

PythonAirflowPostgresDocker

Explore the planned architecture

Start with Turkish discovery seeds; preserve locale information for future languages.

Inside the Model

INSIDE THE MODEL

One question. Multiple signals.

V1 architecture design

Selected entities from the documented model. Choose a table to explore its grain and the decision behind it.

questions 1 → N mappings N ← 1 search_termsquestions 1 → N matches N ← 1 videosvideos 1 → N metric snapshots
WHY THIS SHAPE?
questions

Model the viewer’s question independently from the words used to search for it. Merge phrases only when the viewer expects substantially the same answer.

Immutable raw payloadsStaging & enrichmentSemantic modelsDashboard

Design principles: preserve missing evidence as unknown; record AI classification provenance; serve the dashboard only from semantic models. This diagram describes the intended V1 architecture, not a claim that every component is implemented.

Read the case study

The problem

A popular video is a signal, not proof of a repeatable opportunity. Creators need context around demand and competition before choosing a topic.

A deliberate decision

Start Turkish-first while preserving multilingual capability. Use Low / Medium / High confidence rather than suggesting false precision with a numeric score.

Where it stands

The repository contains the Docker Compose, Airflow and Postgres foundation and a small example ETL DAG. Live ingestion, enrichment and scoring are the next implementation steps.

What comes next

Validate a source-to-semantic slice before expanding coverage. Evaluate AI enrichment against reviewed examples before using it in decisions.

02 / 0→1 PRODUCT / ON-DEVICE VISIONTestFlight milestone

Later, Nope.

A reminder isn’t an action.

A hydration app that connects a scheduled break to a brief on-device verification flow. From product decisions to a TestFlight build.

iOSAlarmKitVisionTestFlight

Explore the product loop

Onboarding sets a daily goal and waking hours; the app calculates the schedule.

From Product Decisions to Interface

FROM PRODUCT DECISIONS TO INTERFACE

A little attitude.
A deliberate daily loop.

The approved design reference that guided Later, Nope. Explore the Home, Alarm, Proof and Verified screens.

Cobalt creates emphasis
The alarm and key messages use contrast to draw attention.
Acid green marks progress
Verified breaks get a distinct visual signal.
Short copy sets the rhythm
Direct prompts move the interaction from a reminder to an action.
Product design reference · interactive mockup

Illustrative states, not live app data. The proof screen is a visual reference and does not access your camera.

Read the case study

The problem

Dismissing a reminder does not tell us whether the intended action happened. The product explores a more deliberate completion loop.

A deliberate decision

Keep camera processing on device. Distinguish technical unavailability from a missed break, so system failures are not treated as user failures.

Where it stands

Home, onboarding, contextual permissions, settings, alarm handling, Vision verification and shareable diagnostics. Version 1.0 Build 1 reached TestFlight in August 2026.

What comes next

Validate the distributed build with testers. Investigate verification friction and reliability before making claims about sustained habit change.

03 / DATA INGESTION / DISCOVERY MVPConcept · building next

Career Signals

Go straight to the source.

A proposed discovery tool for data and product roles, collected directly from selected company career pages.

Ingestion designData qualityChange tracking

Explore the proposed ingestion flow

Proposed scope: 20–30 companies and one or two supported source types.

Read the case study

The problem

Company career pages are fragmented. Comparing roles is harder when titles, locations and listing freshness are inconsistent.

A deliberate decision

Prioritize provenance, freshness and deduplication. Discover directly from employers without claiming that a role is absent from every other job board.

Where it stands

This is a scoped concept, not a launched platform. The proposed first audience is people exploring data and product roles in Germany or remote-friendly opportunities.

What comes next

Validate source access and user needs, then build a narrow daily ingestion loop with filtering and CSV export. Add AI classification only if it improves discovery.

02 / THE THREAD CONNECTING IT ALL

I care about what
happens after the pipeline.

Does the data mean the same thing to everyone? Can the next team trust it? Does the product solve a real problem?

My path through quality, compliance, process improvement and data governance shapes how I build today. I connect business requirements with architecture decisions—and stay hands-on through implementation.

PROFESSIONAL PRACTICE · FREENOW

01

Architecture with ownership

Led the Commercial Finance domain migration, from infrastructure and schema design to ingestion pipelines and the consolidation of financial metrics.

02

Data people can rely on

Technical leadership for a business-critical reporting migration, with KPI consistency, validation and downstream reporting continuity at its core.

03

Business context, technical depth

A foundation in data governance and Lean Six Sigma, combined with hands-on engineering in Databricks, AWS, Terraform and Airflow.

TOOLS I BUILD WITH

PythonSQLDatabricksSparkAirflowAWSTerraformData modelingData governance

03 / WHAT’S NEXT?

Good systems start with
a good conversation.

Let’s talk about data architecture, technical ownership and useful products.

Say hello ↗Explore my GitHub ↗