Pantelis R. Vlachas

Developing generative models for spatial transcriptomics and computational pathology, working with Kalin Nonchev in Prof. Gunnar Rätsch's Biomedical Informatics Group at ETH Zurich.

Research in collaboration with Prof. Eleni Chatzi (Chair of Structural Mechanics) at the intersection of machine learning and computational mechanics, focusing on modeling and forecasting parametric dynamical systems across structural, fluid, and PDE-based domains. The work emphasizes developing data-driven models that remain generalizable, adaptable to varying system parameters, and capable of real-time continual learning.

Work on advanced deep learning–based modeling of multiscale dynamical systems by developing adaptive, interpretable, and generalizable frameworks. Focused on building surrogate models capable of responding to evolving system dynamics in real time and revealing the functional roles of components within neural forecasting frameworks. Led the development of AdaLED, an adaptive learning framework that couples surrogate models with physics solvers to deliver accurate, accelerated simulations across previously unseen regimes.

Work on data-driven modeling of complex dynamical systems, combining machine learning with first principles from fluid and molecular dynamics. I developed efficient surrogate models for high-dimensional, nonlinear, and chaotic processes, creating methods that enable fast and stable long-term forecasting and significantly accelerate multiscale simulations. My work introduced new stochastic, memory-based, and reduced-order modeling approaches, clarified the strengths of different recurrent architectures, and delivered substantial speedups, up to several orders of magnitude, while preserving physical fidelity. This research advanced the reliability, interpretability, and computational efficiency of simulation-driven scientific workflows.

Analyzed and optimized ADMM and AMA optimization algorithms for model predictive control by reformulating MPC problems and applying relaxation and scaling techniques to improve convergence.

Worked on optimal control for the DLR anthropomorphic hand with variable stiffness actuators, developing fast braking-trajectory estimation methods and applying techniques such as decoupling and eigenvalue-based analysis to simplify system dynamics. Designed and evaluated alternative braking control strategies, improving overall efficiency and responsiveness of the actuator system.

• Assisted in the delivery of courses including Optimization Methods, Control Systems, and Advanced Control Laboratory. • Supported students in grasping complex mathematical concepts, enhancing their analytical skills. • Developed supplementary materials that improved student engagement and understanding. • Collaborated with faculty to refine course content, ensuring alignment with academic standards at TU München.

• Developed a comprehensive model of the NAO H25 robot during the single support phase using Autolev. • Implemented obstacle avoidance algorithms for the LWR robot through learning by demonstration in MATLAB. • Conducted optimal control design analysis for a double pendulum's swing-up, enhancing understanding of motion primitives. • Performed convergence analysis on learning algorithms for dynamical systems, contributing to theoretical advancements in robotics.

• Developed an innovative automated method for detecting floating nodes and short-circuit paths in analog circuits. • Implemented a constraint programming-based algorithm in C++, achieving significant performance improvements. • Evaluated the algorithm's effectiveness in Cadence Virtuoso, resulting in a speed-up over existing solutions. • Contributed to research published at the IEEE International Conference in Electronics, Circuits, and Systems, Cairo 2015.

Designed and built a custom PCB and digital display control circuit, managing the full workflow from schematic design and manufacturing to testing and debugging. Improved visual output quality and enhanced overall system reliability through iterative validation and refinement.

Interdisciplinary one-year master’s program under the supervision of Prof. Telis Tympas, focusing on the social studies of science and technology. The program explores the intersections of economics, ethics, law, policy, and technology, emphasizing the societal challenges and political economy of technological change. Key themes included the co-evolution of science, technology, and society, the ethical and legal dimensions of emerging technologies, the policy and governance of innovation and sustainability, and the role of technology in shaping modern economies and social structures.

Focus on control theory, embedded systems, machine learning, robotics, and optimization.