I’m a computer engineer focused on applied AI, language technologies, and product engineering—from model experimentation and evaluation to complete web and mobile experiences.
01 / Selected work
Selected work across culture, playful digital products, social experiences, and applied language technology.
Culture, thoughtfully indexed
A curated digital home for discovering art, museums, and the stories that connect them.
Tiny games, instant delight
A cheerful browser playground for quick games, sweet rewards, and playful little challenges.
Mythology around the table
A mythology-infused social deduction game of hidden roles, shifting alliances, and divine intervention.
Historical documents, made legible
A multimodal transcription system for Ottoman Turkish documents, built around measurable model quality.
02 / Approach
I work where models, software systems, and user experience meet.
I’m a computer engineer focused on applied AI and product-oriented development. My work often spans data preparation, model integration and evaluation, backend services, and the final interface—because the seams between those layers are where product quality is won.
A model becomes valuable when it survives real constraints and supports a clear human need.
Evaluation, traceability, and understandable behavior make intelligent systems easier to trust.
The fastest way to learn is to place a coherent version in people’s hands and observe.
03 / More projects
Explainable AI · Open source
A local-first CLI and Python toolkit that turns image-classification predictions into Grad-CAM heatmaps, overlays, JSON, and standalone reports.
Applied NLP
An interpretable multilingual spam pipeline combining fine-tuned DistilBERT probabilities with deterministic scam and phishing rules.
ML tooling · Open source
A local CLI and Python library for detecting label conflicts, duplicates, imbalance, text anomalies, and train/test leakage before training.
Language technology · npm
Open-source Turkish plain-language tooling that identifies formal or bureaucratic phrases and suggests clearer, more accessible alternatives.
Computer vision
Local-first extraction for industrial documents with tables, mixed text, and custom preprocessing.
Enterprise systems
Approval chains, notifications, metadata, and internal document workflows built on Frappe.
04 / Capabilities
I use machine learning as part of complete products, not only as isolated notebooks.
Translate model capabilities into understandable, useful product behavior.
Treat evaluation and error analysis as product work, not an afterthought.
Carry a system from prototype through services and into people’s hands.
Selected tools
05 / Now working on
06 / Contact