Get To Know More About

Shouvik Sengupta

Occupation

Data Scientist

Education

B.Tech. Electrical and Electronics Engineering
M.S. Data Science

Data Scientist with expertise in advanced analytics, machine learning, and software engineering. Skilled at transforming raw data into actionable insights and communicating technical concepts to diverse audiences. Seeking a challenging role to apply advanced data science techniques, programming skills and problem-solving abilities to build data driven business solutions.

Explore My

Experience

Virufy

Data Scientist

Sep 2024 - present

Developed and deployed AI models on AWS for COVID Detection from coughs for early diagnosis using audio files
Researched and adapted technologies to build and scale seamless cost-effective ETL pipelines reducing processing time by 5%

University of Colorado Boulder

Data Science Course Assistant

Sep 2023 - May 2024

Mentored, graded and developed tests for Machine Learning course, fostering supportive learning experiences for 60 students

Tata Consultancy Services

Software Developer

Jun 2021 - Aug 2022

Orchestrated the migration of Restful microservices from Python to Java Springboot, optimizing the performance by 20%
Optimized performance by 10% by implementing Kafka based messaging and MongoDB for data storage and backup
Utilized Mockito for unit testing, and deployed the project seamlessly using docker containers over Amazon web services
Automated build, testing and deployment to AWS using Jenkins, facilitating CI/CD workflows for robustness and scalability

Schneider Electric Infrastructure LTD

Intern

May 2019 - Jul 2019

Built Arduino-based rack-in rack-out counter for switchgear, reducing cost by 50% compared to its mechanical counterpart
Participated and delivered presentations, and automated updating of weekly FMEA excel and powerpoint using Python

Browse My Recent

Projects

Driver Profiling

The project focuses on developing models using CNN-LSTM and XGBoost to analyze telematics data, leveraging isolation forests to detect anomalies, assess driving patterns, and score driver safety. Interactive Tableau dashboards made insights accessible to non-technical audiences, supporting safer driving practices and decision-making.

COVID Detection

The Chest X-ray Pneumonia Detection project involved deploying an AWS EC2 web application powered by deep learning to identify pneumonia from chest X-rays with 87% accuracy, supporting medical diagnosis. The workflow incorporated tools like MLflow, DVC, GitHub Actions, and modular coding to ensure efficient experiment tracking, version control, and seamless CI/CD integration.

Conversation emotion-cause pair extraction

The Conversation Emotion-Cause Pair Extraction project focused on modeling the extraction of emotion-cause pairs from conversations. By leading a team and employing effective delegation and goal setting, the project achieved 61% accuracy in emotion classification through fine-tuning a RoBERTa model with PyTorch, enhancing expertise in NLP. Transfer learning was applied to a BERT SQuAD Hugging Face model, attaining a 62.2% proportional F1 score in causal span detection.

Statistical Analysis of Greenhouse Gases

The Statistical Analysis of Greenhouse Gases project involved using regression modeling to forecast CO and NOx levels, providing actionable insights for emission management. The work included exploratory data analysis (EDA), data cleaning, feature selection, and diagnostic techniques, resulting in a 6% improvement in the adjusted R², enhancing the model's predictive accuracy.

English Premier League Predictor

The English Premier League Predictor project involved building an ETL pipeline with Airflow and AWS to ingest and store EPL data in Amazon Redshift for efficient SQL-based extraction and analysis. Machine learning models were trained to predict match outcomes, and Power BI dashboards were developed to support data-driven decision-making.

Anime Recommendation

The anime recommendation web application, built using Flask, featured a recommendation model based on content-based filtering. Anime data was web-scraped, and user reviews were classified as positive or negative using BERT sentiment analysis, enhancing personalized user experience.

Shouvik Sengupta

Occupation

Education

Skills

Programming Languages

Python

R Studio

SQL

Matlab

Frameworks

TensorFlow

Scikit-Learn

PyTorch

PySpark

Flask

Cloud / Data

AWS

Azure

PostgresSQL

MongoDB

Airflow

Kafka

Other Tools

PowerBI

Tableau

Docker

Git

Experience

Virufy

Data Scientist

Sep 2024 - present

University of Colorado Boulder

Data Science Course Assistant

Sep 2023 - May 2024

Tata Consultancy Services

Software Developer

Jun 2021 - Aug 2022

Schneider Electric Infrastructure LTD

Intern

May 2019 - Jul 2019

Projects

Driver Profiling

COVID Detection

Conversation emotion-cause pair extraction

Statistical Analysis of Greenhouse Gases

English Premier League Predictor

Anime Recommendation

Contact Me