Swapnil Patil Email & Phone Number

Pune

Pune, Maharashtra, India

Demonstrated hands-on experience with problem-solving, data cleaning, data visualization, statistical modeling, and model interpreting.Build and train a Decision Tree Classifier and visualize it graphically.

1) Built a web UI layer with ML model deployed at backend that detects sentiment of students in online classroom using live video from webcam and provides real-time aggregated feedback to instructors about class. 2) Employed essential image preprocessing techniques such as image augmentation, Pixel brightness transformations, gray scaling and standardized images to overcome data imbalance problem and improve image quality for better prediction. 3) Experimented using multiple pre-trained models such as DeepFace, Simple CNN and Multi-Task CNN to reduce training time and improve accuracy, the final model built on Multi-Task CNN achieved class average accuracy of 74%. 4) Deployed quantized model on Heroku using streamlit API that detects sentiment of students in an interval of 5 minutes and when implemented would help instructors adjust their teaching strategies accordingly and improve teaching efficiency.

1) Built a clustering model to identify major customer segments on demographic and psychographic-based data to develop tailor-made marketing campaigns. 2) Engineered features to obtain new features such as RFM, RFMGroup, and RFMScore for getting more details about the customer's purchasing behaviour.3) Implemented K-means on RFM data, evaluated optimal clusters using Silhouette score, Elbow method and Dendrogram and experimented hyperparameter tuning using KNN. 4) Visualized and validated inter-cluster separation and intra-cluster similarities and when implemented can have great effect on customer management.

1) Developed Classification model to predict the ten-year risk of Coronary Heart disease based on different features that will enable proactive actions to avoid risk. 2) Performed missing value imputation using mean, outliers treatment, feature selection using SelectKBest method and implemented Random-over sampling to oversample the minority class observations to overcome class imbalance.3) Implemented Gradient Boosting model over Random Forest, experimented with hyperparameter tuning technique such as Random Search and Grid Search to achieve Recall score of 93% and improved by 10%. 4) Leveraged the concept of Feature Importance plots to understand high risk factors to prevent the cardiovascular morbidities and mortality in the future.

1) Developed a regression model to predict hourly demand of bikes to maintain stable supply, accessible to public at right time and reduce public waiting time. 2) Designed the data pipeline to work on, performed Exploratory Data Analysis and understood the impact of features such as time of the day, weather condition, holidays. 3) Tested various hypotheses, Performed feature selection using Pearson Correlation, encoded categorical features, checked multicollinearity using Variation inflation factor and engineered important features for the model. 4) Built Lasso Regression and XGBoost model, experimented with hyperparameter tuning technique such as Grid Search and Bayesian Optimization and achieved improved Adjusted-R2 score of 94% using XGBoost.

Bachelor Of Engineering - Be, Mechanical Engineering, 7.96/10

Activities and Societies: Executive - Mechanical Engineering Students Association Coordinator - Impulse 2019 National level Technical.

Diploma, 77.45% Passed

Activities and Societies: Cricket Team Captain for Division Matches

Matriculation, 83.40