Hrishikesh Logo Image
Hrishikesh Thakur
hamburger menu hamburger menu close

Hey!, I'm Hrishikesh Thakur

Inquisitive soon-to-be CS graduate with a years of software development experience, eager to leverage coding and analytical skills to make a significant impact.

Resume

About Me Here you will find more information about me, what I do, and my current skills mostly in terms of programming and technology

Get to know me!

Hey there! I'm Hrishikesh Thakur, a passionate student pursuing my Masters in Computer Science at the University of Southern California, with a keen interest in software engineering, data analysis, and machine learning. Originating from SRM Institute Of Science And Technology, my journey has been fueled by a strong foundation in Computer Science, alongside practical experiences in various domains.

As I delve into academia, I'm actively seeking opportunities to apply my expertise in software engineering, data analysis, and machine learning. My experiences at HighRadius Technologies have equipped me with valuable skills in streamlining deployments, developing test automation frameworks, and crafting AI-powered solutions. Beyond the screen, you'll find me exploring scenic routes, revving engines, and embarking on thrilling outdoor adventures. As an avid gamer, I relish the immersive worlds and strategic challenges offered by gaming. I'm excited to leverage my diverse interests and skills to collaborate on exciting projects and make meaningful contributions. Let's connect and explore how we can merge technology with passion to create something truly extraordinary!

Contact

My Skills

Languages

Python
C
C++
Javascript
Java

Databases

MySQL
MongoDB

Technologies

React
Nodejs
WebdriverIO
JSON
HTML
CSS

Libraries

Numpy
Pandas
Scikit-learn
Pytorch

Tools and Systems

Git
Jira
REST APIs
AWS
Jenkins
Windows
MacOS
Agile

Projects Dive into a curated showcase of my technical explorations and accomplishments.

Named Entity Recognition with Deep Learning

In this project, I developed a neural network model for Named Entity Recognition (NER) using the CoNLL-2003 corpus, incorporating a bidirectional LSTM to understand text sequences and GloVe embeddings for nuanced context capture. I enhanced the model's precision with hyperparameter tuning and a CNN layer for character-level insights, achieving an F1 score of 89%.

Project

Part-of-Speech Tagging with Hidden Markov Models

Leveraged Hidden Markov Models for POS tagging in the Wall Street Journal corpus, enhancing model understanding with a tailored vocabulary and rare word handling. Trained the HMM on derived emission and transition probabilities, achieving notable accuracy. Applied greedy and Viterbi algorithms for efficient tagging, reaching a 94.81% accuracy on development data.

Project

Sentiment Analysis

In this project, I analyzed Amazon reviews to decode customer sentiments using ML and deep learning, preparing data for binary and ternary analysis and employing TF-IDF and Word2Vec for feature extraction. Models like SVM, Perceptron, and advanced FNN and CNN were developed, demonstrating the enhanced accuracy and insights neural networks bring to sentiment analysis.

Project

Machine Learning model for Forensic Interview Q&A

• Collaborated with Prof. Thomas D. Lyon on a DataFirst initiative project to develop a machine learning model for automated question type coding in forensic interviews
• Leveraged RoBERTa to fine-tune the model using a dataset of 349,033 real forensic interview questions.
• Multi-Class Classification between four crucial question types: Non-Questions, Option-Posing Questions, Directive Questions, and newly introduced Invitations

Project

Bookpro

• Developed a full-stack web application using React, Express, Node.js, MongoDB, and AWS for CRUD operations.
• Implemented OAuth to enable secure authenticated access to client assets without compromising initial login credentials.
• Integrated Redux for centralized application state management and Redux Offline for handling API calls and persisting state even when the user is offline.

Project

Exploring Complex Regularization Techniques on Image Classification and Sequence Labelling

This project explores regularization techniques for image classification, focusing on L2 Regularization, Dropout, Data Augmentation, and Ensemble Regularization (Dropout + L2) using the Tiny-ImageNet dataset. The findings indicate that L2 and Dropout effectively reduce overfitting and improve generalization, with the combination of L2 and Dropout (Ensemble) providing the best results. Additionally, Data Augmentation significantly enhances model robustness. This work underscores the importance of regularization in developing robust machine learning models.

Project

Contact Have a question or an interesting project in mind? Let's connect! Fill out the form below and I'll get back to you as soon as possible.