Hi there 👋

I like to train DL models and design hardware in Verilog :)

Hit me up if you want to talk tech or code..

Check out my work on

1) Reduced Instruction Set Computer (RISC)

The design and implementation of an instruction set of a stripped down RISC CPU in Verilog. Check it out at Ln11211/Reduced-Instruction-Set-Computer

2) Implementation of Level 1 Data Cache

Implementation and performance profiling of two data cache organizations: Direct Mapped cache and Fully Associative cache with and without random replacement. Check it out at Ln11211/L1-Data-Cache

3) FADE : FPGA Accelerated Deep-learning Engine

An implementation of a FPGA accelerated Deep learning engine for running deep-learning models. Check it out at Ln11211/FADE

FADE is a hardware designed to utilize the Deep learning processing unit (DPU) to run DL algorithms for edge applicaitons.

4) Design and Synthesis of a Stopwatch on a FPGA board

This is an attempt to design and synthesize the hardware circuit of a stopwatch with the funcitonality to "start", "stop", "lap" and "reset" usign Vivado design suite. I'm still working on it and I would appreciate any feedback on it Ln11211/Stopwatch

Deep Learning Projects

1) DHaRT : DeHazing in Real Time

A Deep neural network approach to dehazing in real time. Check it out at Ln11211/DHaRT

DHaRT is a UNet based deep neural network architecture trained with a custom loss function to perform dehazing in real time

2) Sig2Sig: An OFDM fading channel estimator and equalizer

An OFDM Channel estimator and equalizer based on the Pix2Pix architecture.Check it out at Ln11211/Sig2Sig

This is a Deep learning approach to solving the fading channel estimation problem in OFDM communication system, where I train and test an image to image transaltion network and benchmark it against the LSE method

3) Image compression using Neural networks

An attempt at compressing Images using Machine learning algorithms and Neural Networks. Check it out at Ln11211/Image_compressor

where I try simple algorithms such a K means cluster and some neural network arichtectures using Conv blocks and LSTM blocks to arrive at a Image compressor baselined against JPEG algorithm.

4) Other Neural Network projects

These are some NN models on datasets and problems suitable for DeepLearning and Computer vision tasks. Please find the notebooks with code on loading the dataset and working on them in the neural-networks repository

5) Regression analysis projects

These are a collection of projects and kaggle notebooks on datasets suitable for regression analysis that I performed myself. These include multiple datasets and visualisations of the data correlations and the regression fits. Please find them at Kglnotebooks

6) Handwritten Digit-Classifier App

This fun to make, end to end project on the very famous MNIST dataset is a great learning experience for me on deploying ML models that I found in the Google Developer's course. Do check it out at Ln11211/Handwritten-Digit-Classifier-App.

It has always been my goal to deploy ML models on edge devices.

I like cats, they are silly.