Building adaptive intelligence and high-performance algorithms.
I specialize in neuroplastic Adaptive Neural Controllers, Reinforcement Learning, and Neuroevolution. With over a decade of experience across fusion energy and industrial automation, I design tiny, robust AI systems for embedded deployment and edge computing.
Featured & Past Projects
HEX:RESISTANCE
A strategic mobile game featuring a custom-built minimax AI engine. I engineered the opponent AI from scratch, optimizing alpha-beta pruning and heuristic evaluation to run smoothly on constrained mobile devices without native system crashes.
Read MoreMuZero RL Trading Agent
Deep reinforcement learning agent for learning trading strategies based on MuZero for a pre-startup mini hedge-fund in the US. Huge potential for automated trading.
Options Volatility Prognostics
Developed a solution using Temporal Fusion Transformers (TFT) for predicting future volatility of stock options, achieving a 64% win rate on validation datasets.
Fusion Deployment Robotics
Parametric and generative design of long reach manipulators at UKAEA, including custom physics engine implementation and topology/dimension synthesis.
ITER VVPSS Control System
Managed and delivered the full project lifecycle for the control system of the ITER Vacuum Vessel Pressure Suppression System tooling.
ESS Active Cells OPC-UA
Developed an OS-agnostic OPC-UA client library and integrated EtherCAT drivers into the CorteX framework for the European Spallation Source.
Automated Cocktail Machine
Freelance project integrating a Yaskawa robot and Allen Bradley PLC to create a fully automated bartending machine for night clubs.
4kW Subsea VSD
R&D at Nebb on a subsea Variable Speed Drive (VSD) for Brushless DC Motors designed to operate at 3000m depths.
Voith Paper Fabrics SCADA
Designed and commissioned Wonderware and Citect SCADA systems across plants in Sweden and the USA, integrating with Nebb's Process Pilot MES.
Yaskawa Nx100 Interface
Created an interface at Johnson Matthey for 16 Yaskawa robots to receive recipe parameters directly from the PLC/SCADA, drastically cutting setup time.
Experience
Senior AI Controls Engineer
- Developing and deploying Adaptive Neural Controllers (ANC/ANS) based on neuroevolution and neuroplastic neural networks for real-time control of motors, drones, and industrial assets.
- Targeting extremely compact, high-intelligence-per-kB controllers suitable for embedded and edge hardware.
- Leading system identification, training, verification, and testing using digital twins, domain randomisation, and multi-scenario simulation.
Senior Robotics Research Engineer
- Led research on rapid concept design of long-reach remote maintenance manipulators for fusion and constrained environments.
- Designed and implemented a custom physics and kinematics engine to simulate long-reach serial robots, moving from manual design to search-driven concept exploration using MuZero, MCTS, GA, and PSO.
- Shaped the Deployment Robotics research topic within the Magnetic Fusion Research Programme.
Machine Learning Engineer
- Built a Temporal Fusion Transformer (TFT) pipeline to predict stock option volatility, achieving ~64% win rate on validation.
- Integrated TFT into a MuZero-style trading agent, learning policies directly from noisy market data via multi-process data generation and training pipelines.
EUROfusion Engineering Grant Fellow
- Awarded a EUR 250k grant to develop condition monitoring and prognostics methods for the DEMO fusion power plant.
- Designed unsupervised and self-supervised ML approaches (adversarial autoencoders, InfoGAN, Transformers) for anomaly detection and degradation tracking with biased data.
Control Systems Software Engineer
- Led the ITER Vacuum Vessel Pressure Suppression System (VPSS) tooling control system through design, implementation and integration.
- Developed an OS-agnostic OPC-UA client C++ library and integrated a Hilscher EtherCAT driver for the European Spallation Source (ESS) Active Cells project.
Automation Software & PLC Engineer
- Developed and commissioned SCADA systems (Wonderware, Citect) for offshore, energy, and high-volume manufacturing environments.
- Reduced automation-related downtime from ~10% to ~0.7% per month by redesigning critical PLC logic.
- Developed SCADA interfaces for Yaskawa Nx100 robot parameter downloads, improving changeover efficiency.
Education
PhD in Intelligent Machines / Machine Health Monitoring
Research and development of novel Unsupervised Machine Learning techniques for Machine Health Estimation and Prognostics applied on the DEMO future fusion power plant.
MSc in Automation, Robotics and Systems Engineering
Thesis: Sensor fusion of Inertial sensors and Global Positioning System using an Extended Kalman Filter on a custom mobile platform (Arduino + BeagleBone Black, Ubuntu).
BSc in Computer Systems Engineering and Automation
Graduation project: Algorithm for 2D terrain mapping using a single smartphone camera on a mobile robotic vehicle, combining frame-to-frame triangulation and encoder odometry.
Expertise & Skills
Adaptive AI Control
Building neuroplastic, neuroevolution-based Adaptive Neural Controllers (ANC). Crafting tiny AI fibres for embedded control, digital twins, and real-time domain randomisation.
Machine Learning & AI
Deep learning for anomaly detection and prognostics using CNN, LSTM, ResNet, and Transformers. Specialized in unsupervised and self-supervised learning methods.
RL & Neuroevolution
Developing reinforcement learning agents and search-based planning algorithms for robotics and quantitative trading applications.
Robotics & Industrial Systems
Long-reach manipulator concept design, kinematics, and constrained workspaces for fusion reactors. Deep industrial automation experience.
Selected Publications
A Kinematics Optimization Framework with Improved Computational Efficiency for Task-Based Optimum Design of Serial Manipulators in Cluttered Environments
N. Petkov, O. Tokatli, K. Zhang, H. Wu and R. Skilton
2025 IEEE International Conference on Robotics and Automation (ICRA), Atlanta, GA, USA
Remote handling strategy and prototype tooling of the ITER vacuum vessel pressure suppression system bleed line valve assembly and rupture disk assembly
Keelan Keogh, Chang-Hwan Choi, David Cooper, Steven Craig, David Hamilton, Stewart Hockley, James Kent, Chris Lamb, Nikola Petkov, Andrew Robbins, Paul Talbot
Fusion Engineering and Design, Volume 153, 111485
Abstract: As part of the development work performed under the ITER Robotic Test Facility (IRTF) program, development and substantiation of the maintenance strategy for the Vacuum Vessel Pressure Suppression System (VVPSS) must be investigated... During removal the confinement function of vessel must be retained. The remote handling of these components while maintaining the first confinement boundary is challenging due to the access restrictions. To understand and develop the strategy, prototype tooling in a mock-up environment has been created.
Towards rapid concept design of long-reach remote maintenance manipulators using Monte Carlo Tree Search neural networks
Poster J-588 at SOFE 2023
View Poster ListingSelf-supervised deep representation learning adversarial autoencoder: Deep domain adaptation for data-based fault diagnostics
UKAEA preprint UKAEA-RACE-PR(23)03, submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence
UKAEA Scientific PublicationsCost-benefit analysis of condition monitoring on DEMO remote maintenance system
Fusion Engineering and Design 160
Read Publication