MSc Intelligent Systems and Robotics (Distinction) · University of Essex, United Kingdom
Robotics Software Engineer at RB-Ware, building ROS 2 software for industrial collaborative robot cells from the ros2_control hardware driver to the operator interface. Mechatronics & Robotics Engineer specializing in intelligent and autonomous systems, industrial automation, computer vision and machine learning. I design robotic platforms with ROS 2, C++ and Python, build Transformer-based AI models, optimize multi-object tracking algorithms, and ship production AI systems used by 113+ active users across 10 countries.
Combining control theory, artificial intelligence and hands-on engineering to build systems that actually run
I am a multidisciplinary engineer working at the intersection of robotics, control systems and artificial intelligence. My MSc in Intelligent Systems and Robotics at the University of Essex focuses on autonomous platforms: ROS 2 architectures, Gazebo simulation, SLAM and mapping, sensor fusion with Kalman/EKF estimation, and advanced control from PID and fuzzy logic up to LQR and Model Predictive Control. My MSc thesis, SkyCoat, is a fully autonomous high-rise facade painting robot developed through a complete systems-engineering lifecycle - requirements, mathematical modelling, mechanical CAD, ROS 2 software architecture, physics simulation, fault injection and quantitative evaluation.
On the AI side, I develop deep learning systems end to end. I built a multimodal emotion recognition pipeline fusing Vision and Audio Transformers with actor-disjoint validation, and I optimized the ByteTrack multi-object tracking framework with Hungarian matching and an improved Kalman filter, roughly doubling processing speed without meaningful MOTA loss. As founder of CryptOn Forecast, I designed and deployed an LSTM-based autonomous trading and market intelligence platform on AWS that currently serves 113+ active users across 10 countries. Alongside this, my industrial experience at Turkish Aerospace Industries, Mercedes-Benz Türk and in PLC-driven process automation grounds everything I build in real-world integration, validation and safety.
At RB-Ware I work on the software that sits between a collaborative arm's control box and its motion planning stack. The core of that work is treating the hardware boundary as a fault-prone channel: a 500 Hz ros2_control loop that must stay deterministic through network loss, controller resets and operator stops, with every safety-relevant setting re-asserted after each state transition instead of assumed. Collision detection is framed as a residual-detection problem on joint motor current, where the decision threshold is shaped by the commanded acceleration so that model-predicted current changes are not mistaken for contact. Every change is verified first against an emulated controller and then on the physical cell.
ROS 2, Gazebo, SLAM, Sensor Fusion, LQR, MPC, PID & Fuzzy Control
PyTorch, TensorFlow, Transformers, LSTM, Semantic Segmentation
Siemens TIA Portal, S7-1200, Structured Text, HMI/SCADA, Profinet
STM32, FreeRTOS, Raspberry Pi, FPGA, C/C++, I2C/UART
The engineering toolkit behind my robotics, automation and AI work
From aerospace test engineering and industrial PLC automation to industrial robot software and production AI systems
Developing ROS 2 software in C++ and Python for industrial collaborative robot arms in robotic welding and grinding cells, across the full stack from the low-level hardware driver to the operator web interface. Audited and hardened the ros2_control hardware interface, resolving more than 35 issues across freedrive mode, E-stop detection, halt reporting and safe initialisation speed, each verified on the physical cell. Built a robot-free test bench that emulates the robot controller protocol and runs the real driver and joint trajectory controller against it, enabling old vs. new regression testing of failure scenarios that cannot be triggered safely on hardware. Moved collision protection into the driver: control box collision settings and tool payload are re-applied after every stop, reset and reconnect, and payloads refused by the controller are detected instead of silently lost. Added a per-joint collision stop on windowed motor current steps at 500 Hz, with a threshold that scales with the change in commanded acceleration, separating replans and hard reversals from real impacts without losing sensitivity in steady motion. Characterised the limits on the cell for sensor and grinder end effectors and exposed them through runtime services and a web UI panel. Corrected self-collision status decoding, made automatic tool changes work end to end in Gazebo, and ran the hardware acceptance tests for collision, payload, network-loss recovery and the operator UI using touch-event UI automation and rosbag-based drift analysis.
Founded and developed an AI-powered algorithmic trading platform combining deep learning, quantitative finance and real-time market analytics. Designed and deployed a fully autonomous Binance Futures trading system using LSTM-based neural networks, 16+ technical indicators and automated risk management. Built scalable cloud infrastructure with Python, Flask, AWS, REST APIs and real-time data pipelines, alongside a crypto intelligence platform featuring 40+ live analytics tools, on-chain analysis, whale tracking, liquidation heatmaps and market scanners. Now in production, serving 113+ active users across 10 countries with 24/7 automated trading and a secure non-custodial architecture.
Developed a fully automated multi-station laboratory robotics system with six independently controlled magnetic stirring modules under PID-based closed-loop speed and temperature control. Implemented embedded control software in Python on Raspberry Pi, integrated pH, temperature and RPM sensors with real-time feedback, built a Flask/React web interface for remote monitoring, tuned PID parameters in MATLAB, and contributed across the full cycle from prototype design to integration, validation and performance testing.
Worked on automation and process optimization for an industrial wastewater recovery and Reverse Osmosis treatment facility. Developed PLC control software on the Siemens S7-1200 platform in TIA Portal using Structured Text and Ladder Logic for multi-step process sequences, TON timer-based state transitions, analog signal processing, interlocks and alarm management for pumps, valves, motors and filtration units. Analyzed SCADA screens, P&IDs and electrical schematics, reverse-engineered the complete RO process into a deterministic control algorithm, and validated the sequence through simulation and live observation. Hands-on with SINAMICS V20/G120 drives, SIRIUS soft starters and contactors, SM I/O modules and HMI systems.
Gained practical experience in test engineering, PID tuning, LabVIEW programming, calibration and position/velocity control, deepening my understanding of force and torque control in high-precision aerospace applications.
Contributed to a project applying TensorFlow and OpenCV to improve the efficiency of dial error detection on Mercedes-Benz Tourismo buses, producing actionable data for maintenance and repair. The work reinforced the value of AI in predictive maintenance and industrial quality control.
Developed a real-time embedded system on the STM32F412G Discovery board using C, HAL libraries, FSM design and FreeRTOS. The system integrates multiple sensors (ALS LTR-308, MC3479 accelerometer), an LCD-based UI, UART debugging and LED/button hardware via shift registers. A multi-tasking architecture with three independent FreeRTOS tasks enabled sensor acquisition, graphics rendering and human-machine interaction with precise timing and no blocking delays.
MySQL · HTML · CSS · Electronics · Automation · C/C++
Electronics · Automation · OMRON PLC Systems
Siemens Combustion Turbines · Industrial Systems
Academic journey in mechatronics, intelligent systems and robotics
Certified language skills and international qualifications
Quantitative precision at scale - an AI-driven trading system running in production
Driven by multi-factor models, statistical arbitrage logic and a deep learning inference engine leveraging 16+ real-time technical signals, CryptOn dynamically allocates margin across long/short hedged positions - optimizing for risk-adjusted return. The platform now serves 113+ active users across 10 countries.
LSTM-based neural networks processing 16+ technical indicators in real time for optimal trade execution
Hedging strategies with dynamic margin allocation and automated risk management for capital efficiency
Continuous 24/7 market monitoring on AWS with instant position adjustments and non-custodial security
Robotics, computer vision, deep learning, embedded systems and mechanical design
Designed and validated an autonomous facade-climbing robotic system for high-rise coating, integrating semantic segmentation, closed-loop coverage planning, wall-plane state estimation, motion control and fail-safe interlocks. The project followed a complete systems-engineering lifecycle: requirements, mathematical modelling and mechanical CAD through ROS 2 architecture, Gazebo physics simulation, fault injection and quantitative evaluation.
Designed a hybrid Fuzzy Logic + PID controller for autonomous navigation using ROS 2 and a 2D LIDAR. The system performs obstacle avoidance, right-edge following and adaptive corner turning with dynamic direction-locking to eliminate oscillations. Fuzzy membership functions, linguistic rules and centroid defuzzification were implemented from scratch, with PID control integrated for precision wall-following and validated both in Gazebo and on real hardware.
End-to-end multimodal emotion recognition system combining Vision Transformers and Audio Transformers in a late-fusion architecture. Developed the complete preprocessing, feature extraction, training and evaluation pipeline with actor-disjoint validation, Log-Mel spectrogram processing, ViT embeddings, Class-Balanced Focal Loss and real-time inference - improving generalization across unseen subjects through balanced multimodal learning.
Optimized the ByteTrack multi-object tracking framework by redesigning the data association stage with the Hungarian matching algorithm and improving the Kalman filter prediction pipeline. Profiled computational bottlenecks, optimized matrix operations and evaluated multiple tracking approaches, achieving roughly 100% processing speed improvement across several benchmark datasets without meaningful degradation in MOTA performance.
Fully autonomous Binance Futures trading system built on LSTM neural networks, 16+ technical indicators and automated risk management, paired with a crypto intelligence platform offering 40+ live analytics tools, on-chain analysis, whale tracking and liquidation heatmaps. Deployed on AWS with real-time data pipelines and a secure non-custodial architecture serving 113+ users in 10 countries.
Fully automated laboratory robotics system with six independently controlled magnetic stirring modules under PID-based closed-loop speed and temperature control, integrating pH, temperature and RPM sensors with real-time feedback and a React.js/Flask interface for remote monitoring.
Fully interrupt-driven reaction time measurement system on the STM32F412G Discovery board using FSM design and hardware timers, with an 8-LED countdown, random stimulus generation and 1 ms precision LCD display.
Multilayer perceptron built from scratch in pure Python (2-8-2 architecture) predicting thrust and turn commands for a simulated lunar lander, trained via backpropagation with momentum and achieving 0.012 MSE across all datasets.
Deep learning system for detecting human sounds and localizing people trapped under rubble, combining CNN and LSTM architectures with MATLAB-based signal analysis.
Low-power solar energy harvesting micro-system with capacitor-based storage and 555 timer threshold detection, enabling repeatable burst-mode actuation under severe power constraints, fully simulated in NI Multisim.
Clock-driven digital dice system combining an NE555 astable timer with a CD4017 Mod-6 counter and diode-based decoding, validated in NI Multisim and translated to a 2-layer FR-4 PCB layout.
Precision BLDC motor achieving 90.2% efficiency with a 24-slot, 20-pole design optimized for minimal torque ripple and effective thermal management.
Comprehensive building automation system built on Panasonic PLC hardware with PID control loops, motor management and interlocked safety logic.
Industrial vibration detection and analysis using image processing and SIFT feature matching for predictive maintenance applications.
11 kW industrial press drive system with flywheel energy storage, optimized gear transmission and high-reliability bearing selection.
AI-powered autonomous waste sorting with spectral analysis, IoT monitoring and blockchain-based rewards, aligned with UN Sustainable Development Goals.
Manual crane system for 350 kg loads with St52 steel construction, a power screw mechanism and a 3.5 safety factor verified through structural analysis.
Complete 3D CAD design and modelling of a precision robotic arm system, including kinematic layout and component selection.
Wireless controller designed around PIC microcontrollers with RF communication, custom CCS C firmware and a dedicated PCB layout.
System identification and PID controller optimization using root locus techniques in MATLAB for improved transient performance and stability margins.
Download my complete professional resume
The complete overview of my professional experience, education, projects and technical skills.
Good things, chosen by hand - our small family run online store
A small, family owned and run store. Every item on the shelf was picked, checked and packed by us, and we keep the catalogue deliberately small - only things we would happily use in our own home make it in. Card payments are handled end to end by iyzico, so customer details never touch our servers.
Open to robotics, autonomous systems, control and AI engineering opportunities