SELECTED PROOF FOR CLIENT REVIEW

Systems I have actually built—and the boundary of my ownership.

Real work is separated from planned capability. Sensitive client and customer data is never exposed.
01

Independent project · 2023

Selfiedreams / Alicem Generator

Problem

Personal image models needed an interface that friends could use without operating the inference stack directly.

What I built

Python/Streamlit interfaces with prompt input, text-to-image API requests, generation parameters and display of decoded image outputs.

My ownership

Personal model experiments, interface development and prompt-to-image API integration. This is a historical project, not a current hosted inference service.

Outcome / evidence

Original August 2023 source code and generated images preserved in the project archive.

PythonStreamlitREST APIGenerative AIPillow
02

Co-founder · Product & Engineering

Portföy Uzmanı

Problem

Private property inventory and live client demand were fragmented across messages, spreadsheets and personal networks.

What I built

A permission-aware PropTech network with property and professional profiles, role-based access, demand records, matching, operational dashboards and AI-assisted workflows.

My ownership

Product architecture, database model, frontend product system, matching logic, AI features and production delivery.

Outcome / evidence

182 verified professionals in five weeks with ₺0 advertising, as of 21 August 2026.

Next.jsTypeScriptSupabasePostgreSQLMapLibreLLMsPostHog
03

Hepsiburada · Data Scientist · Growth → Search/NLP

Search & Ranking Intelligence

Problem

Customers use incomplete and ambiguous language; the system must reliably rank the most useful products.

What I built

NDCG evaluation, a 20K-item human relevance workflow, reranking features, category scoring, brand popularity, typo/query work and experiment dashboards.

My ownership

Data analysis, modeling and feature development, relevance evaluation and experiment measurement.

Outcome / evidence

Four years of work across relevance, ranking, recommendation and query understanding.

PythonSQLBigQueryNLPTransformersRankingExperimentation
04

Independent work · client details sanitized

AI Operations Systems

Problem

Leads, documents, conversations and decisions lived in separate systems and depended on manual follow-up.

What I built

Lead-to-CRM workflows, property sourcing and extraction, communication automation, company/document intelligence and operational dashboards.

My ownership

Discovery, data and workflow design, APIs, structured extraction, AI behavior, monitoring and handover.

Outcome / evidence

Reusable patterns for moving unstructured work into measurable operational flows.

Pythonn8nMakeREST APIsCRMRAGLLMsAnalytics