At TÜPRAŞ, I architect and provide technical leadership for an equipment digital twin platform that forecasts equipment behavior and detects anomalies. It runs on 9 assets across 3 refineries and has flagged developing faults 8 hours ahead of alarms.
I scaled catalyst monitoring from a pilot to 20+ units across 3 refineries, giving process engineering teams forecasts and dashboards to plan turnarounds earlier. I also built a closed-loop MLOps framework for product-quality soft sensors, with drift monitoring, automated retraining, and versioned rollback.
I'm currently designing an LLM agent that retrieves evidence from equipment manuals and live process data, then drafts source-cited root-cause hypotheses and recommended actions for operator approval. Earlier at TÜPRAŞ, I delivered 8+ AI and data projects and developed a maintenance scheduling solution that cut planning time by 90%.
My earlier work included back-end services for YouEtix / Passaparola and digital twin, NLP, and computer vision projects at Ind Software. I've also co-authored peer-reviewed work published by Springer, SciTePress, and MDPI, and contributed to EU-funded R&D.

