Vadim Demchik Email & Phone Number

Leipzig, Sn, De

Leipzig, Saxony, Germany

Lead Developer and Architect for Yocto-Based Embedded Software Platform. Designed and developed an advanced embedded software platform for Senec Home4/P4 home energy storage systems (successfully launched in Q3 2023).Failure-Safe Embedded Platform Development* Engineered a failure-safe embedded platform tailored for home energy storage systems, prioritizing robustness and reliability.* Delivered the energy management core, DC-AC and DC-DC-PV drivers, and other critical modules using C++, Yocto, STM32, CAN-bus, and IPC.Product Stabilization and Performance Optimization* Led stabilization efforts post-launch, ensuring optimal system performance and reliability.* Conducted comprehensive battery and system data analytics to enhance operational efficiency and performance monitoring. System Health Monitoring* Remote health monitoring analysis for large-scale deployment, managing data from over 200K home energy storage systems and 700K+ batteries containing dozens of cells. Utilized Databricks, dbt, SQL, Spark, Python, PowerBI, and Kafka to build a scalable infrastructure for continuous system health analysis.Battery State Prediction Model* Developed a transformer-based neural network (PyTorch, Python) to predict battery states within a two-hour window, enabling precise energy management and predictive maintenance.

Founded hgpu.org, the first free and open GPU-hardware platform, by sharing nodes from a custom-built GPU cluster optimized for HPC research. This initiative established hgpu.org as a globally recognized hub for GPU and HPC computing, attracting leading organizations such as NASA, Nvidia, Intel, and AMD.* Built hgpu.org into a premier resource for GPU-based computing, aggregating cutting-edge research, tutorials, and tools for academic and industrial users.* Fostered a global community of researchers and developers, driving interdisciplinary collaboration and advancing GPU applications in AI, machine learning, and scientific simulations. * Promoted open knowledge sharing, shaping the adoption and evolution of GPU technologies across various fields.* Engaged with key stakeholders in academia and industry, positioning hgpu.org as a trusted authority in HPC.This pioneering platform continues to serve as an indispensable resource for the GPU computing community, advancing innovation and collaboration worldwide.

Leipzig, Germany

Deep Learning and AI SolutionsContributed to ExB's leadership in cognitive computing by developing core libraries and pioneering solutions in deep learning, machine learning, computer vision, and NLP. Delivered scalable, HPC systems and innovative algorithms that advanced intelligent data processing and analysis. High-Performance Computational Libraries* Designed and implemented a highly parallel, self-balancing, out-of-core BLAS library supporting dense, sparse, and half-precision computations for extremely large-scale tasks (C/C++).* Developed a parallel, out-of-core SVD, optimizing computational efficiency for massive datasets (arXiv: 1907.06470).* Created a parallel, out-of-core CRF algorithm, significantly enhancing accuracy for NLP tasks (C/C++).* Engineered parallel ML algorithms, including SVM, KNN/ANN, decision trees, and clustering, optimized for large-scale systems (C/C++). Optimization and Performance Engineering* GPU/OpenCL and multithreaded CPU/C++ systems to enhance scalability and performance in computationally intensive applications.* Delivered expert-level performance engineering, ensuring efficient resource utilization and rock-solid reliability of AI-powered systems.Deep learning and AI Solutions* Applied deep learning architectures, including CNNs, RNNs, and LSTMs, for advanced document processing in computer vision applications (Python, TensorFlow).* Developed models for extracting structures and relationships within complex datasets, enabling more insightful data analysis.* Built a LRP package to interpret and explain neural network behavior, improving model transparency and trustworthiness (Python, TensorFlow).

Dnipro, Dnipropetrovsk Region, Ukraine

Branch Director – Aviso (Dnipropetrovsk)Founded and successfully managed the "Aviso" advertising newspaper in the Dnipropetrovsk region, achieving rapid growth and market leadership. Key accomplishments include:* Building the company from the ground up, overseeing all aspects of operations from strategy to execution.* Managing a team of 50+ employees, fostering a productive and results-driven organizational culture.* Achieving profitability within six months of launch, demonstrating strong leadership and financial acumen.* Establishing "Aviso" as the top advertising newspaper in a region with a population of 3.2M, securing its position as a trusted resource for advertisers and readers.* Driving innovation and operational efficiency to maintain market dominance and profitability.

Researcher – Quantum Chromoplasma Laboratory, Dnipropetrovsk National UniversityConducted groundbreaking research in quantum field theory (QFT), with a focus on investigating phenomena such as the electroweak phase transition in intense external fields. Combined advanced analytical techniques with non-perturbative methods to achieve significant scientific contributions. Key accomplishments include:* Monte Carlo Lattice Simulations: Specialized in applying high-parallelization Monte Carlo methods to explore QFT phenomena, leveraging C, Fortran, and MPI to manage the immense computational demands of these simulations.* Pioneering GPU Computing: Among the first to adopt GPGPU technology (using C/C++, AMD IL, and OpenCL) for scientific computations, achieving breakthroughs in computational efficiency and scalability.* Custom GPU Cluster Development: Designed and built a GPU cluster, later deployed as the first open-hardware platform for GPU computing available over the Internet. This platform became integral to the HGPU Group, fostering global collaboration in high-performance computing.Key Achievements in QFT:* Investigated electroweak symmetry breaking and the spontaneous generation of magnetic and chromomagnetic fields in both the Standard Model and the Minimal Supersymmetric Standard Model.* Conducted extensive studies on lattice Monte Carlo simulations for QFT phenomena, including the phase transition behavior in the 𝜙^4 scalar model.Scientific Software Development:* Created several computational tools and algorithms tailored for QFT research, including massively parallel pseudorandom number generators, Monte Carlo simulation frameworks, and data analysis utilities.This research not only advanced the understanding of fundamental physical phenomena but also contributed to the development of innovative computational techniques and platforms that have been adopted globally.

Doctor Of Philosophy (Ph.D.), Theoretical Physics

PhD thesis: "Research of electroweak phase transition in intense external fields" Supervisor: Prof. V. Skalozub, Dnipropetrovsk National.

Postgraduate, Theoretical Physics

Master’S Degree, Theoretical Physics, Diploma With Honours