Projects
October 2020
– May 2021
Designed a simple FPGA synthesizable (tested using Xilinx Vivado) 32-bit RISC processor with its own ISA in verilog.
The ISA has 2 types of instructions where we can have 2 source registers or 1 source register + Sign extended constant like the R-type and I-Type instructions from RISC-V.
Further implemented a simple 128-bit AES security module in verilog for encrypting and decrypting Memory transactions in the aforementioned processor.
Built a simple high-level assembler (in Python) for supplying encrypted machine code.
The motivation behind this project was to provide protection against hardware probing attacks in low-cost RISC based processors for applications such as IOT.
January 2021
– March 2021
![](https://www.gstatic.com/images/icons/material/product/1x/docs_32dp.png)
Object Recognition using RPI
Identified MobileNet-SSD v2 as the model with a good balance between accuracy and low computational requirements after testing
out multiple object detection models on the Raspberry Pi 4B + Pi Cam.
Stored images of detected people locally to show its viability as an edge-computing intrusion detection system.
Setup SFTP(remote file) access to the raspberry pi for secure remote access to the stored images via Filezilla on any PC.
Course Project under Dr. Sumam David, NITK
March 2020
– April 2020
Cleaned up empty/missing values and identified relevant columns in a large loan approval dataset (500k+ samples, 20 columns) using techniques like PCA, Correlation matrix to prevent overfitting to the dataset.
Tested out relevant machine learning models on the dataset after understanding their working.
Identified XGboost as the highest accuracy model on the validation dataset after tuning models with over 60% accuracy.