Ankur Deshmukh

- Title: “Hardware Compiler for RISC-V compatible Accelerators targeting AI Applications” - Develop a compiler, which can translate neural network models described using Tensorflow into hardware accelerators for fast prototyping - The accelerator is ASIC-synthesizable and scalable based on available resources on FPGA or area constraints for ASIC implementation - The accelerator contains AHB-lite interface for interfacing as a co-processor

- Part of the InnoRetVision project group. - Designed hardware accelerator for a neural network used for prediction of the different retinal cell activations based on stimuli - Platform used: Xilinx Arty A7 for accelerator testing and Raspberry Pi for generating stimuli.

Title: HW acceleration of Depth Estimation Techniques for edge computing using FPGA Description: Implement monocular Depth Estimation based on deep learning on FPGA and use for real time inference for better performance on edge devices with limited capabilities.

Revamp Hardware Software Co-Design 2 lab. Update HW from Spartan 6 to Zynq 7000 SoC Use latest Xilinx SW tools like Vivado and Vitis. Upgrade the lab demonstrator to use Robotino 3, omniwheel robot

Working as SW developer for autonomous driving functions Responsibilities: 1. Refactor the existing C++ codebase for various optimizations 2. Create helper scripts in Python to visualize the performance of feature based on various code changes done by the team. 3. Refactor codebase to resolve various AUTOSAR/MISRA-C guideline violations

Model Based Systems Engineering for Automotive Systems using SysML and implementation using MATLAB/Simulink. Deep Learning for Real time object identification for ADAS

Using Model Based System Engineering to model the start-up sequence of a hybrid vehicle. The task included modeling start-up components of ECUs and communication between them. The modeling was done on the basis of various documents which outlined how each ECU communicated and which signals are transferred. The final model was simulatable inside the Cameo environment.

Current average grade : 1.79 Specialisation in System Design. Projects : FSR and Myoware sensor based bracelet design Subjects : Design of Digital Systems, Design of Heterogenous Systems, Hardware Software Codesign 1 and 2, Design of Software for Embedded Systems