Rajarshi Paul
rajarshipaul2249@[Link] | +91 8017260701 | LinkedIn
Experience
Data Scientist, Finarb Analytics Consulting Pvt. Ltd. Aug 2024 – Present
AutoML Platform Development (Internal Product) Aug 2024 – Jan 2025
• Collaborated in designing and developing a scalable AutoML platform for automated machine learning workflows.
• Implemented preprocessing, imputation and dimensionality reduction techniques that improved downstream model
accuracy by 8–12% on average.
• Implemented knowledge distillation techniques compressing ensemble models by 70% with less than 2% accuracy loss.
• Improved model f1-score by 12% through integration of robust outlier detection and undersampling methods.
Pricing Optimization – DirectMed Imaging (Client Project) Feb 2025 – Jul 2025
• Analyzed sales data to uncover pricing patterns across modalities, manufacturers, and customer segments.
• Built an R-Shiny dashboard showing item level pricing trends, win rates, and historical bid insights.
• Designed a pricing simulation tool with dynamic waterfall charts and configurable markups based on urgency, product
condition, sourcing status and also engineered similarity-based logic to surface comparable products.
• Conducted customer segmentation using RFM analysis and derived item-level behavioral insights to inform pricing and
marketing strategies.
Projects
Modeling Patient Stay Duration in Hospitals May 2023 – Jun 2023
• Executed comprehensive data preprocessing and feature engineering to improve model readiness and input quality.
• Utilized Recursive Feature Elimination (RFE) for selecting the most relevant features and reducing model complexity.
• Trained and evaluated models (Logistic Regression, Random Forest, XGBoost) using ROC-AUC and other metrics.
• Performed hyperparameter tuning to enhance model performance, achieving 94% accuracy with XGBoost.
A Predictive Study of Cancer Mortality Across US Counties Oct 2022 – Nov 2022
• Conducted diagnostic tests to validate assumptions of linear regression.
• Utilized Weighted Least Squares to address heteroscedasticity, boosting R² from 0.52 to 0.84.
• Identified and removed influential data points using Cook’s Distance to enhance model robustness
• Detected multicollinearity using Variance Inflation Factor (VIF) and addressed it through LASSO regularization
• Evaluated predictive performance of LASSO and stepwise regression, achieving an adjusted R² of approximately 0.89
Image Classification on CIFAR dataset Sep 2023 – Oct 2023
• Built and trained a Convolutional Neural Network (CNN) for classifying images in the CIFAR dataset.
• Improved model performance by increasing the number of filters and convolutional layers.
• Applied Max Pooling to reduce dimensionality and mitigate overfitting.
• Used ReLU activation to introduce non-linearity and Softmax for multi-class prediction
• Achieved 88% accuracy after systematic hyperparameter tuning.
Education
• [Link] | ISI Kolkata Aug 2022 – May 2024
• [Link]. in Statistics | IIT Kanpur Jul 2021 – May 2022
• [Link]. in Statistics | RKMRC Narendrapur Jul 2018 – Jul 2021
Skills
Statistical Analysis • Time Series • ANOVA • Hypothesis Testing • T-test • Chi-square test
Machine Learning • Regression • Classification • Clustering • Ensemble methods • Reinforcement learning
Deep Learning • CNN • RNN • Word Embeddings • Transformers
Tools & Technologies • Python (pandas, numpy, scikit-learn, statsmodels, tensorflow, matplotlib, seaborn) • SQL • R
Credit Risk Modeling • Logistic Regression • PD/LGD/EAD Modeling • Weight of Evidence • IV
Interests • Risk Analytics • Deep Learning • NLP • Generative AI
Coursework • Google Data Analytics • Machine Learning • Deep Learning Coursera
Languages • English • Bengali • Hindi