About Me
I lead engineering on a greenfield SaaS for robot fleet operations — backend, data model, auth, ETL, deploy pipeline — and run our internal AI tooling effort, pairing agentic coding with project-specific context.
I’m at my best on small teams making real trade-offs. I prefer architecture decisions that serve the product, not the diagram — stacks sized so one engineer can hold the whole system in their head when an incident hits.
- Location
- Istanbul, TR
- Education
- CS undergrad — İstanbul Ticaret Üniversitesi
- Languages
- Turkish (native) · English
Experience
Fullstack Engineer
2025 — PRESENT
Kora Analitik — greenfield SaaS for robot fleet operations · Istanbul
- Took the platform from an empty repo to a working pilot — backend, data model, auth, ETL, deploy pipeline. Now running for pilot customers and live demos.
- Kept long-running workflows consistent when robot connections dropped — reconciling partial state, retries and recovery were the hard part, not the happy path.
- Integrated the MQTT broker with the application backend — reconnection, message ordering, per-tenant topic isolation.
- Containerised the platform with Docker and owned it past the deploy — per-tenant isolation and data security as first-class concerns, not bolt-ons.
Python (FastAPI · Flask · Django) / MSSQL / Redis / MQTT / Docker / React
Works
Robot Fleet Operations Platform
PILOT · LIVE
Greenfield SaaS taken from an empty repo to a working pilot — scheduling and monitoring long-running robot workflows that survive partial failure.
- Problem
- Fleet orchestration that only handles the happy path falls over in the field.
- Approach
- A deliberately focused stack — FastAPI, MSSQL, Redis, MQTT — with retries, recovery and per-tenant isolation designed first.
- Result
- Empty repo to working pilot — now running for pilot customers and live demos.
FastAPI + MSSQL + Redis + MQTT + React
Portfolio Intelligence
IN PROGRESS
A finance side-project: analytics and insight over investment portfolios — ML over multi-layered data processing, with strict per-tenant isolation.
- Problem
- Portfolio data lives in layers that are hard to read as one signal.
- Approach
- ML over multi-layered data processing, with data security as a first-class concern — not a bolt-on.
- Status
- In progress — data pipeline first, insights on top.
Python + ML + React
kernelp4n1c
IN PROGRESS
A self-hosted personal site and multi-author blog engine with a deliberately retro-terminal identity — kernel-panic branding, amber-on-black CRT UI, scanlines and a blinking-cursor prompt.
- Problem
- Own the whole publishing stack — content, auth, media, infra — instead of renting a hosted blog/CMS, without the UI turning into generic template soup.
- Approach
- Decoupled Astro frontend against a JSON API, Markdown rendered and sanitized server-side (syntax highlighting, auto TOC, allowlist), MinIO for media and Redis for cache and background jobs — containerized, single-node Compose with a path to k3s.
- Status
- In progress — working end-to-end (2FA auth, admin CMS, threaded comments, live Spotify/GitHub/WakaTime widgets, RSS); CI/CD and observability still open. First versions launching soon at kernelp4n1c.com.
Spring Boot + Flyway + Redis + MinIO
Internal AI Tooling
INTERNAL
CLI-based agentic coding with custom-built workflows, custom RAG and knowledge graphs — plus internal training on agentic dev practices.
- Problem
- Agentic coding without project context produces generic code.
- Approach
- CLI workflows wired to custom RAG and Obsidian knowledge graphs.
- Result
- Runs as the team’s internal tooling effort, with training on agentic dev practices.
Claude Code / RAG / Obsidian
Stack
- Backend
- Python (FastAPI, Flask, Django) · MSSQL · PostgreSQL · Redis · REST APIs · MQTT · Kafka
- Frontend
- React · Next.js · Astro · Vue
- AI & Tooling
- Claude Code · Codex CLI · custom RAG · knowledge graphs (Obsidian) · agentic dev training
- Infra & Testing
- Docker & container management · end-to-end testing · data integrity
- Fluent in
- Java / Kotlin · C / C++