About

I build AI systems that run in production.

I am an AI Engineer at Ata Teknoloji Platformları (ATP, Ata Holding). I write the backend of an endpoint automation platform that manages tens of thousands of Windows and Linux devices - the kind of system that resolves faults without human involvement.

The centre of gravity is agentic AI: LLM-based agents reach internal systems through defined contracts, RAG runs over internal documents, and inference runs on the company's own servers. Data never leaves.

Measured results

Figures come from the platform's own dashboards.

21–27K incidents / month resolved without human involvement
37% autonomous of the operations team's workload
10–20 min → 1–4 min average resolution time
10,000+ devices managed Windows and Linux endpoints

What I built on the platform

Backend architecture

Split the backend into single-purpose microservices with defined inputs, outputs and failure behaviour. Job scheduling, retry, rollback and fault tolerance live inside the system; retry count, backoff and rollback point are set per task type.
Agent systems

I build multi-agent systems on OpenClaw and NVIDIA NemoClaw. Agents reach internal systems through tool calling and function calling contracts - through verifiable interfaces, not free text.
Agent evaluation

Every agent step is validated against a schema. I monitor task success through tracing and LLM-as-a-judge evaluation, and keep human-in-the-loop where mistakes are costly.
On-premise inference

Self-hosted LLMs on Ollama and vLLM. Embeddings, vector search and RAG over internal documents. Company data never leaves the network at any stage.
Self-healing and orchestration

Built the layer that runs PowerShell and Bash across device groups, and the layer that restarts services and repairs configuration from health checks. Incidents flow into ServiceNow without changing the IT team's existing process.
Security

Prompt injection defence and sandboxing. I version API contracts for backward compatibility; input validation, session and key management, OWASP practices.

Skills

Artificial Intelligence: Agentic AI · Multi-Agent Systems · LLM · RAG · Embeddings · Vector Search · Tool & Function Calling · Human-in-the-Loop · LLM Evaluation · MCP · Prompt Engineering
AI Tooling: OpenClaw · NVIDIA NemoClaw · Ollama · vLLM · LangChain · LangGraph · LlamaIndex · Qdrant · Pinecone · Weaviate
Backend: Python · Node.js · TypeScript · NestJS · Express · REST APIs · Microservices · API Versioning · Software Architecture
Data: PostgreSQL · SQL · Database Design · Migration · Query Optimization · Indexing
Automation & Operations: IT Automation · Endpoint Management · ServiceNow · Job Scheduling · Fault Tolerance · Self-Healing · Structured Logging · Monitoring
Platforms & Security: Linux · Windows · PowerShell · Bash · On-Premise Deployment · Application Security · Prompt Injection Defence · Sandboxing · OWASP