Stan Maciag

• Improve end-to-end inference efficiency and stability for large-scale GPU serving. • Ship optimizations across kernels, compilers, and runtime to lower latency and cost. • Co-design models and systems with platform teams.

• Supported MTIA LLVM-based kernel and low-level DAG compiler team. • Defined the compiler roadmap and shipped multiple milestones from early dev hardware through tape‑out support. • Delivered mid‑ and low‑level optimizations that improved perf/TCO and met kernel performance goals across multiple chip‑generations. • Supported next‑gen MTIA ISA definition, pre‑silicon validation, end-to-end compilation enablement, performance scoping, experimentation, and optimization.

• Led the integration of NPU hardware accelerators into DirectML/Windows OS, driving industry AI computing advancements. • Managed a team specializing in machine learning frameworks and compilers, ensuring seamless releases of NPU DirectML with Intel AI PC. • Delivered key inference performance benchmarks for ISVs, strategizing for scaling across new hardware generations.

• Directed the AI Compiler Optimization team within the OpenVINO framework. • Managed a multinational team for critical NPU IP programs. • Developed engineering roadmaps for precise execution timelines and resource allocation.

• Led the Neural Network compiler project at Intel's Movidius division, driving key performance improvements. • Orchestrated stakeholder engagement and defined project scope across interdisciplinary teams. • Designed a new NPU compiler from scratch, delivering to external customers. • Drove team execution and roadmap, ensuring successful project delivery.

• Orchestrated the creation of an automated HVAC software prototype with advanced heat regeneration capabilities. • Developed a multifaceted hardware/software system with actuators and sensors for control automation. • Designed electronic control circuits for HVAC automation, including the control module and user interface.

• Developed remote motion control software for the Oncilla robot in the AMARSi Project. • Engineered an on-board server on the Pigeon RB100 single-board computer using Linux Ubuntu. • Utilized C/C++, Linux API, and Xenomai framework for server development. • Crafted a companion desktop client application in Java for seamless server communication.

Major - Robotics Thesis: topic - Automatic module for tracking of moving object in mobile robotics, implementation of the object tracking algorithms (Matlab, C language), design and implementation of complete tracking engines for use in the mobile robotics

Thesis - topic: Architecture of the vision system for mobile robot, design and implementation of the stereo-vision system with ability to recognize objects and compute their relative position (C++, OpenCV)