Let me introduce myself
Knowledge is Power
Things I do for love
You’ve got a friend in me
My name is Prasannakumaran. I am currently in my senior year in Computer Science and Engineering at SSN College of Engineering. A passionate computer science student with great interest in Machine Intelligence and Human behaviour. Highly motivated to build solutions for real world problems which could help the community. I love to learn and spread my wisdom and have lead several projects and our research work got published in conferences in Australia and Poland. You can find me in late night DotA pubs, building my militart base in Age of Empires or exploring the world of Witcher when am not working. I have a great passion for gaming and computer technology. I Have played numerous games since I was 5 and even built my own custom pc in my early teens. My other interests include binge watching tv shows, movies and anime. My favourite fictional characters include Eren Yaegar (Shingeki no kyojin), Batman, Jamie Lannister (Game of Thrones) and Michael Scott (The Office).My current works are on Explainable Machine Learning and AI for health.
I'm currently studying towards a Undergraduate degree in Computer Science and Engineering at SSN College of Engineering. I've spent 2 years at Solarillion Foundation in Chennai, where I got introduced to Machine Learning and Research. SSN is a renowned institute in the country with qualified faculties and rich alumni network. I got interested with graph based learning and did multiple projects at College and Solarillion Foundation. I intend to pursue my Masters in Computer Science in Fall 2022.
2004 - 2018
High School Courses
Bachelor of Computer Science
2018 - 2022
Research & Teaching Assistant, Server Maintenance and Development
I have been a Teaching Assistant at Solarillion since March 2020. I have mentored over 15 students during their orientation phase. Have taught and helped the students clarify doubts in Python and Machine Learning and also guided them during their machine learning project phase where they learnt data wrangling, modelling and analyze results. Am an active member of the project review commitee and have reviewed the project reports of over 20 students. I started my journey as a Research Assistant in December 2019. My research work in fake news detection where I lead a team of 5 and my work in malware analysis where I was the second author got published in renowned International conference in Australia and Poland respectively. Further, I am a core member of the server team where I developed modules and maintain the compute server at Solarillion Foundation.
A Non-Intrusive Machine Learning Solution for Malware Detection and Data Theft Classification in SmartphonesMachine Learning Research
In this work, we proposed a non-intrusive machine learning pipeline for malware detection using Machine Learning. Further,
we identify the type of data stolen by the malware. I was the second author in this work and my contributions include
formulating the problem statement, designing the architecture of the machine learning pipeline, data pre-processing, additional
experimentation (progressiv`e learning), analyzing and validating results.
We tested the performance of our proposed model on the publicly available data collection framework -Sherlock which is the largest open-source real-world mobile dataset which holds 4 terabytes worth of data. Our architecture exhibits less than 9% inaccuracy in detecting malware and can classify the type of data that is being stolen with 83% certainty. and Our work got published at ICCS'2021 which is an A ranked conference in CORE classification.
SOMPS-Net : Attention based social graph framework for early detection of fake health newsDeep Learning Research (Graphs)
In this work, we proposed a social-graph based framework for early detection of fake news. Our propsed solutions does not
consider the article content and our work even though aimed towards fake health news detection, the framework can be
generalized across all domains. I lead a team of 5 which consisted of both juniors and senior researchers. My contributions to
this work includes formulation of problem statement, designing the architecture, data pre-processing and analyzing the results
We consider the Twitter engagements (tweet, retweet, replies) and the meta information of the article to classify a given article as fake or real. Our architecture utilizes used Graph Convolutional Neural Networks (GCNN) and Multi-Head Attention (MHA) and its performance was tested on the FakeHealth dataset which contains fake health news on various topics such as cancer, alzheimer's and Stroke. Our model outperfomed other state-of-art fake health news models by 17.1%. The model is capable of detecting fake news with 79% certainty within just 8 hours of its broadcast.
ECMAG - Ensemble of CNN and Multi-Head Attention with Bi-GRU for Sentiment Analysis in Code-Mixed DataDeep Learning Research (Natural Language Processing)
In this work, we proposed an ensemble framework for sentimental analysis of code-mixed text. The work is part of the
Sentimental Analysis for Dravidian Languages in Code-Mixed text task at FIRE 2021. I lead my team under the
guidance of Dr. D. Thenmozhi. My contributions to this work includes designing the architecture and data pre-processing.
Our model was tested on the code-mixed YouTube comments dataset. Our architecture utilizes the XLM-R sub-word embeddings and the two components Convolutional Neural Network for Texts (CNNT) and Multi-Head Attention pipelined to Bi-GRU (MHGRU) utilizes these embeddings and the outputs from the two components are combined and this data is used for final prediction.
Department Graph Database SystemBackend Development, Database Management
I got the opportunity to develop the backend for the department's database system.
The project is aimed towards helping the faculties retrieve records from otherwise complex datatabase system.
We developed a flexible graph-based database system to store and query college data and was built on Neo4j and NodeJS.
My contributions to this work includes designing the schema which generalizes across multiple use cases, designing and developing the two sub databases in which one database handles the CRUD operations and the other handling the permissions for the various entities (nodes) such as Teacher, Event nodes on the graph database. I developed the connector for the system which queries the database using Neo4j API and providing the restructured result to the front end.
I learnt a lot in backend and database management from the project and to write clean and bug free code.
I started this project as a good gesture and giving what I could back to the community.
The idea of the project surfaced during the COVID-19 pandemic where people were disturbed both mentally
The developed Flask web application aims to mitigate these problems and promoto overall well-being of the user. The application uses APIs to create a meal plan for the user based on preferred diet such as Keto, Vegetarian, Paleo etc. There is a personal blog section and news feed section exclusively containing latest health related news. To promote the social well being of the individual we created a discord server and our own subreddit.
We aim to conduct watch parties, game streams and other interactive activities in our Discord server. The application is deployed on the Heroku web server and is currenly live.
TARS: Workplace Automation Bot for Solarillion FoundationDevelopment
I worked on this project during my time in the server maintenance and development team at Solarillion Foundation.
I automated the schedule meeting module. In this module, the meeting is booked, and the people participating in the meeting
is notified and their google calendar is updated with the event and the hangouts meet link. This project proved to be useful
by saving time and is currently used by over 30 people.
I learnt how to work with APIs, integrating the backend and the cloud database. I am currently working on revamping the organization's website and the orientee tracking module. In the future, I am looking forward to work on the automatic assignment grader module for grading the python and machine learning assignments.
Flight Delay Prediction using Machine LearningMachine Learning
I worked on this project during the orientation phase at Solarillion Foundation. This project discusses about the accurate methods involved for classifying whether a flight will be delayed and predicting the arrival delay in minutes using Machine learning. In this work, the historic weather records and flight information was considered for predicting the delay of a flight. I learnt how to work with big datasets, transforming the data, modelling and analysis of results.Languages and technologies used