• Designed and delivered production-ready RAG pipelines using Redis, PostgreSQL, and pgvector, serving real-time semantic search across multi-thousand document corpora
• Built multi-agent agentic workflows orchestrating 4+ LLMs (OpenAI, Claude, and local models) within a single backend system, reducing manual task handling to zero for targeted workflows
• Architected FastAPI-based microservices handling concurrent AI workloads with sub-second response latency under normal load
• Containerized and deployed applications on AWS using Docker, integrating Grafana and Prometheus for full observability across all services
• Shipped 3+ end-to-end products covering AI backend, deployment, and Next.js frontend within solo engagements
• Engineered deterministic system behavior to ensure reliability and maintainability long after delivery

• Ranked 17th out of 32 finalist teams in TEKNOFEST 2023, a nationwide AI competition with hundreds of applicants across Turkey
• Built a real-time YOLO-based object detection pipeline classifying 6+ vehicle categories from drone imagery with live inference
• Implemented UAV ambulance landing zone feasibility logic using spatial constraints and multi-class detection outputs
• Integrated model with official competition APIs for real-time evaluation across multiple qualification stages
• Awarded Honourable Mention — the only solo, high-school-level participant competing among university teams

• Engineered predictive maintenance models (TensorFlow regression & LSTM) across 7 rolling mill motors and 20 vibration sensors, achieving ~89% fault detection accuracy
• Reduced unplanned maintenance response time by enabling early fault detection on 3 months of historical time-series data
• Built and deployed a Flask-based monitoring dashboard visualising real-time machine predictions vs. actual sensor performance
• Designed OpenCV-based steel width measurement tooling used in daily production-line quality control
• Collaborated directly with production engineers to validate model outputs against real operational constraints

Production-ready AI assistant handling product discovery, FAQ resolution, order tracking, and store logic via real-time WebSocket communication. Supports 4+ concurrent tool-calling agents, vector search across product catalogues, and full observability via Grafana dashboards.

A semantic search engine built on 500k cleaned IMDb entries using Qwen embeddings, OpenAI intent parsing and a multi-step ranking pipeline. It understands user preferences such as genre, theme, tone and region to return the most relevant titles.

A full-featured corporate website for a Turkish heavy-load logistics company operating across 13 countries. Includes a custom CMS panel with AI-powered automatic translation, a quote request ticketing system with automated email notifications, a newsletter system, and media storage on Cloudflare R2. Integrated interactive route maps and Google Analytics.

Custom desktop ERP for a local scrapyard. Replaced a fully paper-based workflow that made it hard to track real profit, cash flow, and inventory. Now the client tracks materials, weights, expenses, and daily balance in one place. Used in daily operations since launch.

Completed high school diploma via distance learning while working full-time as an AI Engineer. Transferred from Adnan Oguz Anadolu Lisesi after landing a professional engineering role at age 16.

Lifelong self-taught engineer. Core knowledge built through hands-on projects, official documentation, research papers, and courses across deep learning, LLMs, backend systems, and cloud infrastructure. All skills validated in production environments.