I build AI products that professionals use in their actual workflow.
My background is a double engineering degree — electrical & electronics and energy systems — and, since the end of 2021, a day job on carrier-grade network infrastructure at Türk Telekom. Four years of that work taught me one reflex: when a problem repeats, turn it into code. Most of what I've automated there — device provisioning, 5G synchronization, fault tracking — started as someone's manual routine.
Products came later. Network work is satisfying, but its feedback loop is internal: you make the machine run better and nobody outside ever sees it. I wanted to build things people choose to use. So I started shipping — first MentAI with a small team of close friends, then mulakat.pro on my own — and both eventually stopped when real usage didn't match the design. With MentAI we went through the whole early-stage loop: formed a team, talked to users, sat in founder meetups, learned that distribution matters as much as architecture.
ConsulDent is the third attempt, and it is built differently: with a practicing dentist, inside the workflow it serves, instead of from the outside looking in. That is the main lesson the first two left me — a well-designed system isn't a product until someone's actual day depends on it.
The two tracks feed each other. Telecom taught me to design for failure and to distrust anything that only works in a demo; building products taught me to ask who actually needs a thing before engineering it. ConsulDent gets both. "Workflow" is the word I keep coming back to: a tool that lives outside the workflow gets opened twice, one that lives inside it stops being optional. That is the standard I hold the products to.
Tools
Python · Django · FastAPI · PostgreSQL · pgvector · LangChain / LangGraph · Whisper · Netmiko · .NET · Linux
Experience
2024 — present
A clinical AI assistant for dentists, live on the App Store. Answers are grounded in retrievable clinical sources and returned as structured output that fits the clinic's day: visit preparation, notes, patient communication. I own it end to end — RAG pipelines, backend, voice input, deployment, and the product decisions in between. It is currently in testing with a cohort of around fifty dentists, and payments are live. Full write-up →
Dec 2021 — present
Carrier-grade infrastructure on Nokia and Huawei platforms: MPLS services, access networks, and the operational reality of keeping them running. My focus is turning network problems into reproducible, code-driven systems — replacing repetitive operator work with tooling the team actually adopts.
2023
Completed the data science & machine learning bootcamp, then returned as a teaching assistant for the following cohort — statistics, scikit-learn, model evaluation.
Earlier roles
Earlier projects
Education
2017 — 2021
GPA 3.42 / 4.00 · 100% scholarship
2016 — 2020
GPA 3.72 / 4.00, highest in cohort. Graduation work: a smart home energy demand management system — real-time consumption data driving demand reduction without hurting comfort. IEEE paper →
If you're building something where AI has to work inside a real professional workflow — or you just want to compare notes — I'd like to hear from you. I'm most interested in clinical and other high-consequence domains, where structured, source-grounded output matters more than a chat box.