backend engineer. trying to become a data engineer.
I build APIs, data pipelines, and backend systems designed to survive retries, restarts, and the edge cases that eventually become normal cases.
Mostly Go, Python, PostgreSQL, and Linux.
Rewriting the backend around Django, PostgreSQL, and SeaweedFS. The work covers data modeling, storage boundaries, API behavior, and migrating existing functionality without losing what already works.
I also help with frontend work when needed—the occasional coat of paint.
Django · PostgreSQL · SeaweedFS · backend rewrite · occasional frontend
A pre-publication review system for sponsored content. Video, speech, OCR, captions, evidence, and versioned rules are brought into one review trail.
I am building the backend with one other engineer: a Go modular monolith, several processes, durable workers, and PostgreSQL as the source of truth. The processes share domain packages rather than communicating over internal HTTP.
Go · PostgreSQL · River · durable jobs · versioned rules · object storage
A local-first personal document store built in Go. Original bytes live in a filesystem-backed, SHA-256 content-addressed blob store; metadata and FTS5 search live in SQLite. The goal is local ownership without giving up reliability or recoverability.
Import, search, export, soft delete and restore, tagging, OCR jobs, doctor checks, backup and restore, plus byte-level export integrity verification.
Go · SQLite · FTS5 · OCR · SHA-256 · filesystem blob storage
A student information system MVP built mainly to learn Go and relational database design. It uses layered REST handlers, service-owned business rules, and PostgreSQL-backed repositories.
Course enrollment is transaction-safe: constraints and pg_advisory_xact_lock handle concurrent registration. Triggers, audit logs, materialized views, and GiST exclusion constraints keep consistency where it belongs—in the database. It is a learning project, not a production reference; there are probably mistakes worth finding.
Go · PostgreSQL · SQL · REST · advisory locks · GiST constraints
Slack existed. A separate institutional coordination system was requested anyway. WhatsApp-style work tracking was moved into announcements, assignments, availability, approvals, audit logs, and notifications.
Role-based workflows cover the system owner, general manager, unit manager, staff, and student assistants. The backend includes an append-only audit ledger, notification outbox, scheduled reminders, admin dashboards, and performance testing with Locust and Prometheus.
FastAPI · PostgreSQL 16 · React Native/Expo · SQLAlchemy · Alembic · Docker
First-place NeuroBridge Hackathon project: a drug-screening pipeline built from SMILES inputs for BBB permeability, logBB exposure, and Tox21 toxicity risk. Eighteen classical ML models were trained and compared.
ExtraTrees-based models reached approximately 0.95 AUC in internal evaluation. Probabilistic outputs were combined with medicinal-chemistry rules to provide interpretable risk explanations alongside the predictions.
Python · FastAPI · scikit-learn · Docker · model interpretation
Worked on İÜ-KİK from March to June 2026: backend services, PostgreSQL schemas, integrity constraints, indexes, service boundaries, API behavior, and transaction-safe institutional workflows.
Led seven technical events with speakers and partners including Red Hat, EPAM, PyData, and Turkish AI Community. Coordinated speaker outreach, partner communication, programming, planning, and event-day operations.
Built relationships across engineering, research, open source, AI, data science, machine learning, and neuroscience. Not every useful system boundary is technical.
B.Sc. Computer Science, Istanbul University. GPA 3.65/4.00, expected 2028.
The formal version has dates, commas, and fewer opinions:résumé.