Adithya Kommini
Independent Projects

PCvalueBuilder: A recommender system to assist customers in upgrading their personal computer (PC) hardware components by optimizing value and performance.

Designed a recommender system to assist customers in upgrading their personal computer (PC) hardware components by optimizing value and performance. Built a data pipeline to extract the performance benchmarks and prices of different hardware components to train machine learning models. Implemented a random forest regression model to predict the component performance and deployed the web app using streamlit on AWS.

A convolution neural network model for Spoken digit recognition using Mel-frequency Cepstrum.

Spoken digit recognition using the Mel-frequency cepstral coefficients (MFCCs) and convolution neural networks (CNN). The model can recognize 0-9 spoken digits from .wav audio files by passing their MFCCs as an image input. My model achieves an accuracy of 94% on the test audio files.

Implementation of image classifier for RGB images using artificial neural networks (NN).

Given a set of RGB images (32x32 pixels) with one (and only one) of the following objects: aves, flights, bucks, felines (labels 0, 1, 2 and 3, respectively). The goal is to train a Neural Network model to recognize which of the objects is present in an image. Also, Understand the efficiency, performance, and perform model selection using cross-validation.

Optimal image compression for efficient reconstruction using unsupervised learning.

Unsupervised learning algorithms attempt to learn some structure of the data using unlabeled samples like images. Here we will use K-Means to achieve optimal compression for efficient reconstruction of images.

Air quality prediction using hyperparameter regression analysis.

Train decision trees with different maximum depths, nearest neighbors with different number of neighbors, and linear models with different regularization parameters to predict the air quality.

Design and Analysis of a Feed-forward PUF (32 nm node) under Voltage Scaling.

Analysis on the effect of non-linear delay model (exhibited by Feed-forward PUF) on performance parameters of PUFs i.e. Reliability and Uniqueness under different operating voltage conditions.

Nanoscale Implementation of G-Share Branch Predictor.

Implemented a G-Share branch predictor using N3ASIC, a nanofabric using combination of crosspoint nanowire FETs integrated using metal interconnects. The usage of N3ASIC reduces area and the improves performance of the predictor when compared to the conventional CMOS.

An AES-Based Intellectual-Property Identification in System-on-a-Chip (SOC) Design.

A novel intellectual-property (IP) identi cation using System-on-a-Chip (SOC) watermarking scheme. The principle is embedding di erent Advanced Encryption Standard (AES) encoders in to a System-on-a-Chip (SOC) based watermarking scheme at behavior design level.
[Journal paper][Link]