Daniil Troshkov Email & Phone Number

Abu Dhabi, Abu Dhabi Emirate, United Arab Emirates

• Firmware development of drone tracker for Thingy91 Nordic IoT device with Zephyr run time OS (RTOS) fully secured with Chain of Trust, TF-M and trusted firmware. Updated TLS to 1.3 version in Nordic infra. U-Blox NEO 8M integration to avoid GPS multiplexing. Multithreading programming for LTE/NB-IoT, Bluetooth, Wi-Fi technologies.• SELinux version update and drivers’ debug. Update Linux version from 5.15 to 6.1 and build system from Buildroot to Nix Flakes. Drivers’ peripheral debug… Show more • Firmware development of drone tracker for Thingy91 Nordic IoT device with Zephyr run time OS (RTOS) fully secured with Chain of Trust, TF-M and trusted firmware. Updated TLS to 1.3 version in Nordic infra. U-Blox NEO 8M integration to avoid GPS multiplexing. Multithreading programming for LTE/NB-IoT, Bluetooth, Wi-Fi technologies.• SELinux version update and drivers’ debug. Update Linux version from 5.15 to 6.1 and build system from Buildroot to Nix Flakes. Drivers’ peripheral debug after update, e.g. PCI, USB, tethering, I2C, UART, SPI, NRC7292. HW/DTS debug.• MDM client-server Mesh dynamic initialization and configuration. Zeroconf and NATS connection with events driven multithreading configuration processing on device side and appropriate modification of server side.• ZYMKEY bring up and testing for target device based on Raspberry Pi Compute Module 4 (CM4). Integration zk-libraries (Zymkey module) to customer specific SELinux hardened OS image.• PXE and U-boot net boot integration for target device based on CM4 with SELinux hardened OS. Scripts for multi PXE boot/U-boot net boot for multiple target devices for testing and fast development of new OS images. Show less

Montreal, Quebec, Canada

- Optimization of neural networks from domain specific language (DSL) AI compiler perspectives, which means analysis of neural network structures, development and implementation of algorithms for AI compiler optimizations. In particular, I performed support and development of the open-source Apache TVM machine learning compiler framework for CPUs, GPUs, and machine learning accelerators.- Implementation and debugging of AI compiler optimizations on different levels of neural networks… Show more - Optimization of neural networks from domain specific language (DSL) AI compiler perspectives, which means analysis of neural network structures, development and implementation of algorithms for AI compiler optimizations. In particular, I performed support and development of the open-source Apache TVM machine learning compiler framework for CPUs, GPUs, and machine learning accelerators.- Implementation and debugging of AI compiler optimizations on different levels of neural networks lowering from high graph-execution level till low LLVM-IR and assembler level. Profiling of generated code, estimation of potential improvements and implementation of additional features.- Development of runtime and C/C++ API to run and tuning compiled neural networks on special targets. Deployment and porting runtime and toolchain for specific targets (Raspberry Pi + Pi OS or QNX or Yocto) with user and colleagues support by request. Porting of Deeplite runtime to Blackberry IVY Platform. Prototyping and compiling new AI tasks and running them in specific runtime and targets.- Testing, DevOps duties and Git version support. Development of test coverage for AI compilers from regression testing to performance testing of new and existing optimizations. Tested compiler and runtime to have a baseline for performance and increased it constantly. Used Arm Virtual HW for the testing of different options and different targets. Created complete and extensible infrastructure for building and testing. Merged with upstream and integrated with Travis.- Collaborated with AI research scientists to accelerate and increase performance of next generations of deep learning models.- Mentored and supported junior AI compiler engineers. Show less

Developed a C/C++ cross compiler in order to improve performance of Computer Vision and Machine Learning applications. The LLVM based C/C++(clang/clang++) compiler was ported from scratch to our Digital Signal Processor (DSP) target. C(newlib), C++ (libcxx) libraries and run time libraries were also ported. The solution is a cross platform compiler which hosts in x86 Linux environment and targets DSP-ELVEES baremetal.

Ported the binary compiler. Soft Machines company developed a dynamic binary compiler from ARM target to the Soft Machines CPU (SMI) in order to make the CPU compatible with ARM (ARM-SMI). As a Senior Developer I ran an outstanding project on porting the binary compiler (ARM-SMI) to x86 (x86-SMI). Used QEMU for the automatic comparison of traces and provided the project with tests for the comprehensive coverage of the X86 instruction set. Developed and maintained a region detector of the hot… Show more Ported the binary compiler. Soft Machines company developed a dynamic binary compiler from ARM target to the Soft Machines CPU (SMI) in order to make the CPU compatible with ARM (ARM-SMI). As a Senior Developer I ran an outstanding project on porting the binary compiler (ARM-SMI) to x86 (x86-SMI). Used QEMU for the automatic comparison of traces and provided the project with tests for the comprehensive coverage of the X86 instruction set. Developed and maintained a region detector of the hot code using hardware features (counters) and heuristics to detect the areas of hot code in order to optimize them with a JIT optimizer. The performance has improved and the compilation time has decreased. Show less

Developed and maintained proprietary embedded binary compiler and compiler optimizations for the Elbrus CPU (from x86). Improved performance through conventional optimizations and target specific ones for Elbrus.

Developed a platform for combining data derived from disparate data sources to empower corporate management with accurate, actionable, and consistent information and to facilitate various consolidated reporting (software for banking industry). Contributed to the development of high quality reliable solutions for the needs of computationally challenging industries. Developed relational databases in SQL and Python for financial organizations. Worked both on back end and front end sides… Show more Developed a platform for combining data derived from disparate data sources to empower corporate management with accurate, actionable, and consistent information and to facilitate various consolidated reporting (software for banking industry). Contributed to the development of high quality reliable solutions for the needs of computationally challenging industries. Developed relational databases in SQL and Python for financial organizations. Worked both on back end and front end sides. Maintained data warehousing: customer support. Ensured trouble-free operations. Assist business analysts with the reports by performing SQL queries. Provided customer oriented improvements on demand. Show less