Evam Gilotra

Design and develop FPGA solutions from RTL to software integration, ensuring seamless functionality analysis and validation of FPGA-based modules.

• Developed a PCIe plugin to control DMA(Direct Memory Access) between FPGA DDR4 and Host system for efficient data transfer.
• Developed an IP using Verilog and UVM to be used as a L1 trigger(latency ~15μs) for data acquisition for a Qubit Readout System using Zynq RFSoC based Software Defined Radio(SDR) System. The IP has DMA capability with 95% throughput utilization to transfer the data to PS and PL DDR and a python based software driver to control the IP.
• Cross-compiled an AlmaLinux 8 image for deployment on a Xilinx Zynq UltraScale+, and integrated the IP core along with the corresponding software driver to ensure seamless interaction between the FPGA hardware and the embedded Linux system.

• Completed RTL Design and Verification of feature extraction module with 17 features for a Decision Tree Classifier ASIC(22nm) using Cadence toolchain.
• Optimized the feature extraction module for lower power vs accuracy of the classifier and integrated it with the Decision Tree Classifier Module.

• Completed design and verification of FPGA based prototype to enable PCIe Gen3/4 over photonics ensuring ~100ns latency overhead for disaggregated resource infrastructure for GPU, FPGA, SSD, DDR3/4 etc.
• Extensive use of System Verilog and VHDL for performing functional simulation, debugging, synthesis, timing closure and verification on AMD Zynq and Virtex UltraScale+ in a lab environment.
• Worked towards regular releases while maintaining source code using Git.
• Collaborated with Software teams to resolve bugs in the design and worked towards continuous improvement of the datapath.

• Designed and implemented a hardware accelerator for matrix multiplication (using Verilog and Vivado HLS) to speed up Convolution Neural Network-based Machine Learning algorithms.
• Design was prototyped on Xilinx Zynq Ultrascale+ and integrated with an existing framework to achieve 140x faster results as compared to A53 Arm processor.