top of page
Mail: arkapratimdas70@gmail.com              Phone: +91-6291564985

About Me.

Hi, I'm Arkapratim Das

I’m a data analyst and machine learning practitioner with a background in Bioinformatics and a Master’s from the University of Manchester. I specialize in end-to-end analytics — from wrangling messy data to deploying intelligent systems that drive business value.

Whether it’s risk modeling in fintech, A/B testing in e-commerce, or customer segmentation in telecom and FMCG, I bring both technical fluency and business acumen to every project.

 What I Do Best

  • Transform ambiguous problems into actionable insights

  • Build robust pipelines and predictive models using Python & SQL

  • Design compelling dashboards with Power BI to empower decisions

  • Apply deep learning, natural language processing (NLP), and statistical modeling to real-world challenges

  • Communicate findings clearly to both technical and non-technical stakeholders

Current Focus

I’m currently deepening my expertise in Natural Language Processing (NLP) and continuing to sharpen my skills across the full analytics lifecycle — from exploration to experimentation, modeling to monitoring.

​Tools & Skills

 Programming & Data

  • Languages: Python (pandas, NumPy, scikit-learn), SQL (advanced queries & optimization)

  • Data Handling: Web scraping, APIs, ETL, automation scripts

  • ML & AI: Classification, Regression, Clustering,Time-Series Forecasting, NLP (spaCy, NLTK), Deep Learning (Keras, TensorFlow)

  • Statistical Methods: A/B Testing, Marketing Mix Modeling, Survival Analysis, Hypothesis Testing, CLTV analysis

Business Intelligence

  • Visualization: Power BI, Tableau, Matplotlib, Seaborn

  • Reporting & Dashboards: Interactive dashboards, executive summaries, self-serve tools

Experimentation & Evaluation

  • Experiment design, test/control analysis, model evaluation (ROC-AUC, F1, PR curves)

  • Model interpretation, feature importance, and actionable insight extraction

Soft Skills

  • Storytelling with data

  • Cross-functional collaboration

  • Agile and stakeholder engagement

Projects

SAMPLE VISUALIZATIONS

lloyddash
OVERVIEW
rfm
COHORT
cr
ss
newplot (2)
newplot (3)
newplot (5)

PROJECTS

lloyds-banking-group1426.jpg
lloydcert.png

CUSTOMER CHURN PREDICTION & SEGMENTATION (Lloyds Advanced Data Science Project- Forage)

​

Industry: Banking / FinTech

​

Overview:

Developed a data-driven solution to predict customer churn and segment the customer base for targeted retention strategies. Leveraged transaction data, engagement logs, and demographic features to build a predictive churn model and derive meaningful customer clusters.

​

​

​

 

 

 

 

 

 

 

​​Key Highlights:

  • End-to-end pipeline from raw data ingestion to model deployment artifacts

  • RFM analysis and Demographic Segmentation (K-Means) for business profiling

  • Applied clustering and supervised learning for segmentation and churn prediction

  • Calibrated Random Forest model with optimized thresholding using precision-recall tradeoffs

​

​

Business Potential:

  • Identified 80% of potential churners ahead of time, enabling proactive outreach

  • Supported marketing and retention teams in designing data-backed campaigns

  • Segmented customers into actionable groups like At-Risk, Loyal, Champions

  • Insights used to refine customer targeting, upsell strategy, and retention offers

  • Improved churn model expected to increase customer lifetime value and reduce attrition


Tools: Python (numpy, Pandas, Matplotlib, Seaborn, Plotly, Scikit-learn), SQL, Power BI  

Techniques: Data Wranagling, Feature engineering, RFM analysis, Customer Segmentation, Predictive Modeling.
 

Visualization Gallery

lloyddash
demo - Copy
heatfinal
bicat
binum

Credit Risk Analysis 

CREDIT RISK ANALYSIS

 

 

Industry: FinTech

​

Overview:

​

Analyzed a highly unstructured and noisy credit dataset to predict customer default risk and enable more informed lending decisions. This end-to-end analytics project included data cleaning, feature engineering, exploratory analysis, dashboarding, and predictive modeling.

​

​

​​​

​

 

 

 

 

Key Highlights:

​

  • End-to-end project- Data Ingestion to Deliverables

  • Exploratory Data Analysis to uncover trends and flag risk factors

  • Built a Machine Learning model to classify high risk individuals.

​

Business Impact:

​

  • Enabled the credit team to identify high-risk applicants early in the approval pipeline, improving loan portfolio quality and reducing exposure to bad debt.

  • Provided a scalable, data-driven risk scoring system that supports consistent, unbiased credit decisions — reducing reliance on manual judgment.

  • Supported risk-based pricing strategies by segmenting customers based on predicted probability of default.

  • Delivered actionable insights through dashboards, helping stakeholders understand key risk drivers and optimize lending policies.

  • The model’s improved precision in identifying likely defaulters is estimated to lead to an 18% reduction in delinquent loans, with downstream benefits in cash flow and capital allocation.

Tools: Python (Pandas, Scipy, Numpy, Matplotlib, Seaborn), Power BI, Machine Learning, Dataiku

​

Techniques: Data Cleaning, EDA, Dashboarding, Predictive Modeling, AutoML

cr

Dashboard

Retail Trial Effect Analysis

RETAIL TRIAL EFFECT ANALYSIS: QUASI-EXPERIMENTAL + MACHINE LEARNING ( Quantium Advanced Data Analytics Project- Forage)

​

​

Industry: Retail/ E-commerce

​

Overview:


Evaluated the impact of in-store layout changes on chip sales performance using a combination of quasi-experimental design and machine learning modeling. Delivered a full-stack analytics solution including ETL automation, Power BI dashboarding, uplift measurement, and forecasting.

​

​

​

 

 

 

 

 

Key Highlights:

​

  • ETL & Data Engineering: Built an automated ETL pipeline using Python, SQLAlchemy, and PostgreSQL to clean and structure raw sales data.

  • Exploratory Analysis: Uncovered product trends, top SKUs, and customer behavior; delivered insights on revenue, brand performance, and customer segments.

  • Power BI Dashboard: Developed an interactive dashboard with live database connectivity for real-time monitoring of store and SKU performance.

  • Uplift Measurement: Used quasi-experimental design with matched control stores to measure impact — Store 77 saw a 29% revenue uplift post-treatment.

  • Sales Forecasting: Trained an XGBoost model with lag features and tuned via Optuna to predict baseline sales and validate observed uplift.​

​

Business Impact:​

  • Enabled the retail strategy team to make evidence-based decisions on scaling layout changes by quantifying their impact on key performance metrics.

  • Demonstrated a clear causal relationship between layout changes and revenue uplift using a combination of control-store matching and predictive modeling — boosting stakeholder confidence in the trial's effectiveness.

  • Delivered a real-time Power BI dashboard that empowered decision-makers to monitor product performance and customer behavior during and after the trial.

  • Identified that the layout change led to larger average basket sizes despite fewer transactions, offering key insights for optimizing merchandising and inventory planning.

  • Provided a data-driven framework for future A/B testing and trial design, enabling the business to evaluate store experiments more rigorously and roll out successful strategies at scale.

Tools and techniques: Python, PostgreSQL, ETL, Power BI, KPI development, Machine Learning 

​

Techniques: ETL pipeline, Data Cleaning, EDA, Dashboarding, Control Store Experimental Design, Predictive Modeling.

Dashboard

ss
Screen

E-Commerce Marketing Strategy optimization

​

​

Industry: E-Commerce

​

Overview: 

​

Performed a customer and revenue analysis project for an e-commerce business using RFM segmentation and cohort analysis. The goal was to understand retention patterns, customer value segments, and revenue behavior to guide marketing and lifecycle strategies.

​

​

​

​

 

 

 

 

Key Highlights:

​

  • Designed and implemented RFM segmentation logic to score and group customers by recency, frequency, and monetary value.

  • Created monthly acquisition cohorts and tracked customer retention and revenue contribution over time.

  • Built Power BI dashboards to visualize cohort retention heatmaps, RFM distribution, and product-level revenue insights.

  • Cleaned and pre-processed transaction data using Python and exported analytical outputs for reporting.

  • Developed revenue and retention KPIs to track customer health and lifecycle dynamics.

​

Business Impact:

​

  • Identified that 84% of revenue and 90% of customers came from the UK, supporting a focused regional marketing strategy.

  • Discovery of peak acquisition and revenue in December 2010 guided the timing of future promotional campaigns and onboarding efforts.

  • High retention in November 2011 informed best practices for customer engagement and follow-up strategies post-purchase.

  • Seasonal spikes in revenue and retention during September–October helped inform inventory planning and campaign scheduling.

  • Detection of churn risk among 20% of high-value customers led to the recommendation of targeted win-back campaigns and personalized retention offers.

​

z

Tools: Python, EDA, SQL, Power BI

​

Techniques: Data Cleaning, RFM Segmentation, Cohort Analysis, Dashboarding

Dashboard

OVERVIEW
rfm
COHORT

© 2035 by Site Name. Powered and secured by Wix

bottom of page