Hi, I'm Rakshatha Vasudev
Welcome to my digital footprint!

Working on optimizing body parameters processing extracted from video frames for the one stage solution framework which uses a Component Aware Transformer(CAT) for 3D human body mesh reconstruction.

Worked under Data engineering and analytics platforms in Securities Services Tech and Commodities businuess units.Was involved in design and development of modules for query optimization, job automation, dependent job scheduler and optimizing HDFS access.
Leveraged tech stack : Java , Springboot, SQL, MySQL, Oracle database, Hadoop, Kafka, Hive and Impala, Bash

Worked in Securities Services Unit to develop a log writer for a trading application based on a design pattern solution to log real time application data and to monitor error logs.
Leveraged tech stack : Java , Springboot, Maven, ELK Stack

• Rakshatha Vasudev,Prathmesh Dahikar,Anshul Jain and Nagamma Patil,Sentiment analysis on worldwide COVID-19 outbreak, International Conference on Big Data, Machine Learning and Applications 2021 (Springer).

• Rakshatha Vasudev,Prathmesh Dahikar,Anshul Jain and Nagamma Patil,Epidemic Outbreak Prediction with Ensemble of Deep Learning Models, International Conference on Recent Advances and Innovations in Engineering, ICRAIE 2022 (IEEE).

InsightBot • Implemented a multi-layered RAG-based chatbot at UBHacking to query YouTube playlists. The video transcripts were fetched via YoutubeTranscriptsAPI, utilizing PineCone for vector storage and Llama LLM for generating contextual response.

Epidemic outbreak study using Sentiment Analysis and Predictive Modelling
• Performed sentiment analysis and classification on COVID-19 twitter dataset with BERT to demonstrate the enhanced contextual representation achieving a model accuracy of 95%.
• Devised a deep learning ensemble model trained using the stacked result set of predictions from TensorFlow’s CNN, LSTM and MLP to forecast COVID-19 cases resulting in a robust model with an average r2-score of 0.98.

PlagiaScan • Orchestrated a data structures and algorithm project to determine percentage plagiarism of a text document with an accuracy of 92% using the Rabin Karp algorithm and rolling hash data structure for string pattern matching.

GeoMapper • Developed a GIS-based web application using Leaflet to visualize and analyze spatial data, incorporating file upload functionality to process and display GeoJSON datasets.

PhishNet • Pre-processed over 5 lakh URL records and extracted relevant features to train machine learning models in Pyspark for phishing URL detection.Developed and deployed the solution as a web extension powered by a Flask server.