Ruchitha Uppuluri

Data Scientist

[email protected]

Resume

Hello, I am

Ruchitha Uppuluri.

3

Machine Learning
Projects

1

Python
Project

About Me

I am a Data Scientist driven by a simple belief: behind every dataset is a story waiting to shape better decisions. My journey began with a fascination for how small patterns can change big outcomes, leading me to specialise in Statistical Modelling and Advanced Analytics.

I thrive on transforming complex, imperfect data into insights that drive clarity and efficiency. Whether building predictive models or crafting intuitive dashboards, I focus on solutions that are transparent, reliable, and genuinely impactful. Meaningful progress starts with asking better questions - I’m here to build the data-driven answers.

Key Skills

Python

SQL

Survival Analysis

Machine Learning

Statistical Modeling

My Projects

Healthcare Premium Prediction

Domain/Function: Insurance / FinTech

Study Dropout Prediction in ALSPAC Cohort

Domain/Function: Health Data Science (Longitudinal Cohort Analytics)

Prognostic Modeling of Head & Neck Cancer

Domain/Function: Healthcare / Survival Analysis

Lauki Finance: Credit Risk Modelling (Classification Model)

Domain/Function: Finance / Banking

My Experience

ENT Administrative Assistant – UHBW NHS Foundation Trust, Bristol, United Kingdom

Oct 2023 - Oct 2025

Led data-driven process improvements by identifying bottlenecks in patient follow-up workflows. Developed an interactive scan-tracking dashboard in Excel to visualise patient pathways, resulting in a 30% reduction in follow-up delays.
Enhanced data integrity by validating and cleaning complex patient records. Corrected inconsistencies in the department's database, improving data quality by 15% and ensuring accurate clinical decision-making for multi-disciplinary teams.
Streamlined operational reporting by automating the tracking of scan outcomes. This initiative reduced miscommunication and directly contributed to a 10% increase in patient satisfaction scores by enabling more timely care coordination.

Project Management Intern – Marvell Technology, Hyderabad, India

Sep 2020 - Jun 2021

Optimised project delivery by analyzing sprint performance data. Leveraged analytical insights to streamline task tracking and reporting workflows, boosting overall delivery efficiency by 20%.
Ensured reporting reliability by investigating test-case anomalies in JIRA. Collaborated with cross-functional engineering teams to debug data discrepancies, achieving 100% on-time reporting accuracy for critical deliverables.

Awards & Certificate

Python: Beginner to Advanced For Data Professionals

SQL for Data Science

Math and Statistics For AI, Data Science

Master Machine Learning for Data Science & AI: Beginner to Advanced

Clinical Trial Analysis & Reporting

Let's Connect

Feel free to get in touch with me. I am always open to discussing new projects, creative ideas or opportunities to be part of your visions.

Mail Me

[email protected]

https://www.linkedin.com/in/ruchithauppuluri/

GitHub

https://github.com/ruchitha-meenakshi

Machine Learning Project

Healthcare Premium Prediction

Domain/Function: Insurance / FinTech

An end-to-end machine learning application deployed as an interactive web tool to predict insurance premiums. The model improves estimation accuracy by 18% and enables stakeholders to simulate pricing scenarios in real time, significantly reducing manual calculation effort.

View Project
Engagement

View Project
on GitHUB

Other Tools

Python

Streamlit

Scikit-Learn

Pandas

An interactive user interface allowing stakeholders to input applicant details for real-time prediction.

Final premium prediction output showing the calculated cost based on the input risk factors.

Project Details

Insurance providers often rely on static, manual rule-sets to price premiums, which can lead to pricing errors, slow turnaround times, and operational inefficiencies. We aimed to modernise this process by building an automated, machine-learning-based pricing engine capable of predicting accurate premium costs based on applicant demographics and medical history in real-time.

We built a regression-based pricing model using Python (Scikit-Learn), performing rigorous feature engineering and cross-validation to optimise performance. Crucially, we deployed the model as an interactive Streamlit web application, allowing non-technical stakeholders to input data and receive instant predictions.

The model improved premium estimation accuracy by 18% over baseline estimates, and the deployed tool reduced the time required for manual scenario analysis by 30%, significantly enhancing the speed and consistency of the quoting process.

Machine Learning Project

Study Dropout Prediction in ALSPAC Cohort

Domain/Function: Health Data Science (Longitudinal Cohort Analytics)

A predictive framework designed to identify participant dropout risk for the MRC Integrative Epidemiology Unit. By analysing demographic markers across 14,000+ records, this project provides a reliable heuristic for retention teams to intervene early and maintain study integrity.

View Project
on GitHUB

Other Tools

Random Forest

MICE (Imputation)

Tidyverse

An end-to-end machine learning pipeline designed to handle complex missing data and optimise Random Forest performance.

Domain analysis identifying Demographic and Lifestyle factors as the primary drivers of participant attrition.

Project Details

Longitudinal clinical studies face significant risks to data integrity when participants drop out over decades, creating a need for data-driven methods to identify at-risk participants and prevent data loss.

The goal was to engineer a predictive model to flag likely dropouts among 14,000+ records, enabling the research team to proactively target retention efforts.

We developed a classification pipeline using Random Forests in R, utilising MICE (Multiple Imputation by Chained Equations) to address complex missing data patterns and engineering features based on demographic and behavioural markers.

The final model achieved an AUC of 0.75, successfully identifying the top behavioural risk factors for attrition. This provided the MRC Integrative Epidemiology Unit with a reliable mechanism to prioritise engagement, directly supporting long-term study retention strategies.

Machine Learning Project

Prognostic Modeling of Head & Neck Cancer

Domain/Function: Healthcare / Survival Analysis

A comparative analysis benchmarking Machine Learning against traditional statistics to predict patient outcomes in Head and Neck Squamous Cell Carcinoma (HNSCC). The study utilised 528 patient records from TCGA to identify critical risk factors and improve prognostic accuracy over standard Cox model

View Project
on GitHUB

Other Tools

Survival Analysis

RandomForestSRC

Glmnet

ggplot2

Variable Importance Plot (VIMP) identifying Tumor Stage and Age as the top predictors of patient mortality.

Kaplan-Meier survival curves showing distinct prognostic separation between early and late-stage disease.

Project Details

Head and Neck Squamous Cell Carcinoma (HNSCC) prognosis is notoriously challenging due to variable survival rates that traditional statistical models often fail to capture. To address this, I built a robust prognostic framework using data from 528 patients in The Cancer Genome Atlas (TCGA) to improve prediction accuracy beyond standard Cox Proportional Hazards models.

I conducted a comparative analysis benchmarking traditional methods against modern Machine Learning, specifically implementing Random Survival Forests (RSF) and L1-Penalised Regression (CoxNet) in R to handle censored time-to-event data directly. The final RSF model achieved a C-Index of 0.74, significantly outperforming the traditional Cox model (0.646).

The analysis identified Tumour Stage IVC as the dominant risk factor and demonstrated that ensemble-based ML models capture non-linear clinical interactions far better than linear baselines, providing a validated pipeline for personalised risk stratification.

Python Project

Lauki Finance: Credit Risk Modelling (Classification Model)

Domain/Function: Finance / Banking

An end-to-end Credit Risk Modelling solution built for an NBFC to predict loan default probability and generate credit scores. The project features a full ML pipeline with SMOTE-Tomek imbalance handling, Optuna tuning, and a deployed Streamlit app for real-time decisioning and automated reporting.

View Project
on GitHUB

Other Tools

Python

Scikit-Learn

Streamlit

Optuna

SMOTE-Tomek

Pandas

Live prediction interface showing the calculated Credit Score, Risk Rating, and Probability of Default for an applicant.

Model evaluation metrics showing a 94% Recall and robust separation between good and bad borrowers (ROC-AUC 0.98).

Project Details

Lauki Finance, a rapidly growing NBFC, faced bottlenecks with manual credit reviews, leading to inconsistent decisions and high operational overhead. To address this, I developed an automated Credit Risk Scoring system to predict default probability and assign actionable risk ratings.

I built a complete Python machine learning pipeline, incorporating rigorous data cleaning and domain-driven feature engineering. To handle significant class imbalance, I utilized SMOTE + Tomek links. After benchmarking models like XGBoost and Random Forest, I selected Logistic Regression for its superior interpretability and regulatory compliance.

The final model achieved a 94% Recall for defaulters and an ROC-AUC of 0.98, ensuring high-risk applicants are effectively flagged. The solution was deployed as a Streamlit web application that generates real-time credit scores (300–900) and downloadable PDF reports, successfully modernising the loan approval workflow.