I build backend services, data pipelines and AI agents in Python. My focus is software that solves operational problems: inventory, forecasting, staffing, document processing and monitoring that runs around the clock without anyone watching it.

Before I wrote code, I ran the operations this software is for. I spent eight years leading customer-operations teams of up to 150 people at Turkcell Global Bilgi and KKTCELL. At KKTCELL, I built a national operator's customer-service and CRM operation from scratch: call centre, back office, IVR and the full KPI framework. That background shapes how I engineer:

• I start from the workflow. I model how a warehouse, store or service centre really runs before I design a single table.
• I build it to last. Database migrations, Docker containers, retries and alerting are part of the design, not an afterthought.
• I automate the boring parts. Scheduled services pull data, apply business rules and escalate only what needs a human decision.

Warehouse Inventory API Tracks SKUs, suppliers, stock and purchase orders, and flags low inventory before it becomes a stockout. FastAPI · PostgreSQL · SQLAlchemy · Alembic · Docker	Demand & Staffing Optimizer Forecasts demand for every store and product, then turns it into a staffing plan that meets service-level targets. FastAPI · PostgreSQL · forecasting · Docker	Warehouse Management System Inventory and order API with role-based permissions and JWT authentication. FastAPI · SQLAlchemy · JWT · RBAC
Receipt OCR & Analytics Turns receipts into structured data and serves spend analytics by vendor, category and period. Tesseract OCR · FastAPI · PostgreSQL · Docker	InsightAI Ask a question in plain English and get the answer straight from a stack of PDFs. Vector search · OpenAI · PyMuPDF · FastAPI	Tradewatch Streams market trades in real time and surfaces unusual activity on a live dashboard. WebSockets · Redis · Plotly Dash · Pytest

Private builds · code not public, happy to walk through them

• Personal AI assistant. Local-first, runs 24/7 on my own hardware, with source-linked memory, one-command agents, encrypted storage and strict spending limits.
• Multi-model AI council. Several language models debate a question and return one reviewed answer.
• Unattended monitoring pipelines. A family of GitHub Actions services that watch public data sources on a schedule, filter by business rules and alert by email, Telegram or voice, with retry and outage handling built in.