About
ML Systems Engineer · Research-to-Production AI
I build AI systems for environments where models do not get unlimited time, data, or compute.
My work sits at the intersection of reinforcement learning, agentic AI, control, and distributed systems — with a focus on turning research ideas into reliable production systems for real-world networks.
From control to production AI
My background began in control and optimization, where the central question was how to make systems behave reliably under uncertainty. Over time, that question led me toward reinforcement learning, distributed training, and AI systems that operate under strict latency, reliability, and compute constraints.
Today, I build production-grade ML systems for telecom networks, including AI-native RAN optimization, agentic AI for autonomous networks, and scalable reinforcement learning platforms.
What I focus on
AI for real-world networks
Reinforcement learning and optimization methods for radio access networks, link adaptation, and autonomous control.
Production ML systems
Distributed training, low-latency inference, model compression, evaluation pipelines, and deployment under operational constraints.
Agentic AI
LLM-based reasoning systems connected to optimization and control loops, especially for intent-based network automation.
Research-to-production translation
Taking ideas from papers and prototypes into systems that engineers can run, measure, maintain, and improve.
How I work
Real-system first
I care about AI that survives contact with real systems.
Beyond accuracy
Latency, throughput, system boundaries, and reliability matter as much as model quality.
Operational thinking
Data flow, failure modes, evaluation, and maintainability shape whether AI systems work in practice.
Deployment-aware ML
A good model is not enough if it cannot run where decisions actually happen.
Experience
Master Researcher in Machine Learning
Dec 2023 – PresentEricsson AB, Kista, Sweden
- Architected a distributed multi-agent, multi-objective reinforcement learning platform for 5G coverage and capacity optimization.
- Built distributed training pipelines and simulators for large-scale experimentation and evaluation in 5G environments.
- Led the design of AI-native systems for autonomous 5G operations using reinforcement learning, Bayesian optimization, and LLM reasoning.
- Accelerated AI-native link adaptation from research prototypes to production with compact, efficient models.
- Mentored engineers and advised teams on deployment constraints, system architecture, and long-term ML platform design.
Senior Researcher in Machine Learning
Jan 2020 – Dec 2023Ericsson AB, Kista, Sweden
- Established the commercial viability of AI-native link adaptation through early research, proof-of-concept development, and collaboration with product units.
- Designed scalable training architectures for high-throughput reinforcement learning on HPC infrastructure.
- Reduced distributed reinforcement learning training time by approximately 20× using high-throughput actors, optimized scheduling, and PyTorch RPC.
- Delivered measurable production improvements in 5G RAN systems under real-world latency, reliability, and hardware constraints.
- Co-authored technical papers and presented system-level insights to senior stakeholders.
Development Engineer, Planning & Decision-Making
May 2018 – Jan 2020Scania CV AB, Södertälje, Sweden
- Developed and evaluated robust path-tracking and motion-planning algorithms for autonomous heavy-duty vehicles.
- Improved reliability of planning and control in complex, unpredictable environments.
- Contributed to patent applications in autonomous transport systems through technical documentation and cross-functional collaboration.
- Supervised a master's thesis on real-time motion-planning automation.
Postdoctoral Researcher
Oct 2015 – Mar 2018Paderborn University, Paderborn, Germany
- Conducted research on advanced control and machine learning methods for large-scale networked cyber-physical systems.
- Developed optimization-based decision-making methods to improve the performance of physical systems.
- Contributed to German Research Foundation grant applications on cyber-physical networking.
- Designed and taught a graduate-level course on Control of Mechatronic Systems.
- Supervised multiple student theses.
Education
PhD, Automatic Control
KTH Royal Institute of Technology, Stockholm, Sweden
2009 – 2015
Thesis: Architectures and Performance Analysis of Wireless Control Systems
MSc, Mechatronics Engineering
Istanbul Technical University, Istanbul, Turkey
2007 – 2009
BSc, Control Engineering
Istanbul Technical University, Istanbul, Turkey
2004 – 2009
BSc, Mechanical Engineering
Istanbul Technical University, Istanbul, Turkey
2003 – 2007
Ranked 1st among 155+ students
See the systems behind the story
Explore selected projects that connect reinforcement learning, agentic AI, distributed systems, and production deployment.