Spoorthi G Kunch

Data-focused engineer with experience in telecom operations and strong hands-on expertise in machine learning, NLP, and predictive modeling. Experienced in building end-to-end ML workflows including data preprocessing, feature engineering, model optimization, and evaluation across financial and operational datasets. Proficient in Python and SQL with practical experience applying ensemble models and deep learning techniques to solve structured and unstructured data problems.

Incident Ticket Closure Time Prediction:

• Built regression models to estimate ticket resolution time at the ticket creation stage
• Performed data preprocessing, feature encoding, and handling of structured operational data
• Implemented Linear Regression, Random Forest, XGBoost Regressor, and Neural Network models
• Applied RandomizedSearchCV for hyperparameter optimization
• Achieved R² score of __ and reduced MAE to __ using the best-performing model
• Developed predictive framework to support proactive incident prioritization

Credit Risk Prediction using Ensemble Learning:

• Developed a classification model to predict loan default probability
• Conducted exploratory data analysis and feature engineering
• Implemented Logistic Regression, Random Forest, XGBoost, and Bagging classifiers
• Performed hyperparameter tuning using GridSearchCV
• Achieved ROC-AUC of __ and accuracy of __% using the best-performing model
• Evaluated model performance using precision, recall, and confusion matrix

NLP Text Classification using Word2Vec and LSTM:

• Preprocessed multi-class text dataset using tokenization and lemmatization
• Built TF-IDF and Word2Vec embedding pipelines for feature representation
• Developed Logistic Regression baseline model
• Implemented LSTM-based deep learning model for sequence classification
• Improved accuracy from __% (baseline) to __% using the LSTM model
• Evaluated performance using precision, recall, and confusion matrix

Patient Monitoring System:

• Developed a prototype patient monitoring system integrating health sensors
• Transmitted real-time sensor readings to a Telegram channel using bot API integration
• Implemented threshold-based logic to categorize patient health condition
• Designed system workflow combining hardware inputs with automated monitoring