Verification Work: 1. Tiler GPU Unit. Backend TB - Subsystem Level SystemVerilog. (October 2022 - February 2023). ● Backend - RTL Description: - Part of a Geometry Processing and Binning GPU Core. - Transforms primitives into polygon lists. 2. Texture Mapper GPU Unit. SystemVerilog. (December 2021 – October 2022; February 2023 - present). ● Texture Mapper - RTL Description: - Part of a GPU Shader Core. - Performs texture lookups, filtering, and decompression for GPU shaders.. ● Protocols: - Message Fabric, AXI, ASN, CHP. ● Sign-off 4 projects. ● Team lead for 3 engineers. 3. Streaming Router Block. (October 2021 – 2 months). ● Create vPlan (Streaming Router Block uses several protocols including UART, AST and APB). ● Write Verification Specification document. ● Develop VE using UVM-SV and CRCDV methodology. ● Scrum Master. 4. AMM Controller Block. (September 2020 – 2 months). ● Create Sync Generator vPlan, Sync Interface vPlan and Reset Interface vPlan. ● Write Verification Specification document. ● Develop VE using UVM-SV and CRCDV methodology. ● Debug packet generation and driving. ● Create tests for DUT verification. ● Advanced automation of the verification flow. ● Run regressions. ● Analyze coverage and implement solution to fill it up. ● Write Verification Report. ● Develop the VIPs for educational purpose. ● Project Management was done using Agile Methodology. 5. Demultiplexer Block. (June 2020 – 1.5 months). ● Create vPlan. ● Write Verification Specification document. ● Develop VE using UVM-SV and CRCDV methodology. ● Analyze coverage and implement solution to fill it up. ● Write Verification Report. ● Develop VIPs for educational purpose. Design Work: 1. Accelerating an AI image upscaling algorithm on FPGA using High-Level Synthesis. (October 2020 - 8 months). ● Implement the Fast Super-Resolution Convolutional Neural Network (FSR-CNN) in Python and C++. ● Optimize algorithm implementation for FPGA (reduce execution time by 42X, optimize used area), using real-time coding style and pragmas. ● Convert the algorithm to HDL implementation. ● Synthesize and co-simulate the kernel. ● Verify functional correctness of the Verilog kernel. ● Port the kernel on PYNQ-Z1 FPGA using DMA and Vivado. ● Evaluate the quality of the upscaled images using PSNR – Peak Signal to Noise Ratio. ● Implement the bicubic interpolation algorithm in Python and C++, convert the algorithm in HDL implementation, optimize the kernel and port the RTL implementation on PYNQ-Z1 FPGA. ● Analyze and compare performance profile of the FSR-CNN with bicubic interpolation algorithm. Technologies: Jupyter Notebook, Vivado HLS, Vivado, PYNQ-Z1, FPGA, Convolutional Neural Network, OpenCV, DMA, AXI4-Stream, AXI4-Lite, Pragmas. 2. Parallel to Serial Block. (June 2020 – 0,5 months). ● Implement the Parallel to Serial Block based on the Design Specification document. ● Debug issues encountered during verification. ● Analyze the code coverage. Additional Responsibilities: ● Write AMIQ Blog article - How to Accelerate an Image Upscaling CNN on FPGA using HLS. ● Mentor for 5 interns. ● Mentor for graduation thesis for a student - Develop a FCOV Maximizer using Python Reflection in CRCDV.