Open to opportunities

Soujanya Ghosh

@soujanyaghosh

Message

I build data-driven ML applications, analytics dashboards, and full-stack platforms that turn complex data into action.

Zimbabwe

Message

What I'm looking for

I'm looking for opportunities to apply data analysis, machine learning, NLP, and full-stack development to build practical products and actionable insights.

I've built churn prediction, sentiment analysis, and legal document analysis projects that turn raw data and unstructured text into actionable insights.

For a 478-user churn dataset, I identified 109 high-risk users with an XGBoost model and achieved a 23% churn detection rate through demographic and behavioral feature engineering. I also created Power BI insights to support retention strategies for at-risk segments.

My sentiment classification pipeline reached 88% accuracy using TF-IDF, SVM, and Naive Bayes, with a focus on precision, recall, and interpretable sentiment drivers.

I'm currently developing AI-powered and full-stack products, including a Legal Document Analyzer using NLP, LLMs, LangChain, FAISS, and Streamlit, as well as Engineers Way, a React and Supabase career-networking platform.

Experience

Work history, roles, and key accomplishments

Engineers Way Web Development

Self Employed

Jun 2026 - Present (2 months)

Engineered a full-stack platform using react and supabase to bridge the gap between freshers and experienced professionals for career networking. Designed and implemented a scalable database schema and er diagram to manage user profiles, authentication, and professional networking modules.

React Supabase Database Design API Development Authentication Media Management

Legal Document Analyzer

Self Employed

May 2026 - Present (3 months)

Developed an AI-powered Legal Document Analyzer for contract analysis and document understanding. Implemented clause extraction, document summarization, and legal risk identification using NLP and LLMs.

NLP LLM Faiss RAG Streamlit PDF Processing Vector Search

Churn Prediction Project

Self Employed

Dec 2025 - Present (8 months)

Identified key drivers of customer attrition for a 478-user dataset, providing actionable insights to improve retention strategies. Reduced potential customer loss by identifying 109 high-risk users using an xgboost model with optimized recall.

XGBoost Power BI Feature Engineering Model Evaluation Data Analysis

Sentiment Analysis Project

Self Employed

Jan 2025 - Present (1 year 7 months)

Developed a robust sentiment classification pipeline, achieving 88 percent accuracy by optimizing vectorization techniques like tf-idf. Scaled model performance by evaluating multiple algorithms, including svm and naive bayes, ensuring high precision and recall for text classification.

TF IDF SVM Naïve Bayes NLP Model Evaluation