Bhawna Kaushik

About Me

I'm Bhawna Kaushik — a Data Analyst who gets obsessed with the why behind the numbers before touching a single formula.

Most analysts wait for a task. I start with a business question. Before I build anything, I need to know: what decision does this data need to support? What's broken, and what would fixing it actually cost or save? That mindset is what turns a dashboard into a tool people actually open — and a KPI into something that changes behavior.

In 6 months across two live internships, I've worked across supply chain, hospitality, e-commerce, and ML — not as a learner, but as someone accountable for real outputs:

Diagnosed a 24.83% cancellation rate as the primary revenue leakage driver for a luxury hotel chain — not just flagged it, but traced it to booking channel mix and pricing patterns
Identified Lotus Mart at OTIF 20–22% as the highest churn-risk account in a supply chain operation serving 16K+ orders — and built the dashboard that made that invisible problem visible
Built a CLTV engine (BG/NBD + Gamma-Gamma) that segmented 4,200+ customers by future value — so retention spend goes where it actually matters
Engineered 15+ KPIs across 100K+ customers and 3,095+ suppliers — each one tied to a business outcome, not just a metric

Technically, I work across SQL (CTEs, Window Functions) · Python (pandas, ML pipelines) · Power BI · Excel — end to end, from raw data to stakeholder-ready brief. I've built ML models, automated PDF reporting pipelines, and presented findings to directors in language they care about.

What managers notice: I don't deliver outputs and disappear. I document the logic, anticipate the follow-up question, and make sure the next person who touches my work can understand it in five minutes.

The data always has an answer. I just refuse to stop until I find it.

Key Skills

SQL Power BI Python Excel Machine Learning

Effective Communication Problem Solving Stakeholder Management

My Experience

Data Analyst Intern

Mar 2026 – Present

PurpleMerit (Remote)

Cleaned and validated 1M+ rows across 9+ linked tables, resolving nulls, encoding conflicts, and quality issues prior to modeling.
Engineered 15+ KPIs (delay/return/cancel rates, SLA breach %, RFM, CLTV drivers) across 3,095+ suppliers and 100K+ customers.
Built 4 predictive ML models end-to-end (return prediction, sales forecasting, CLTV, risk classification) from feature engineering to ROI quantification.
Delivered 10+ business outputs per cycle: scored datasets, Excel scorecards, EOD PDF reports, and director-ready presentations.

Data Analyst Intern

Jan 2026 – Mar 2026

AtliQ Technologies (Remote)

Standardized and validated 100,000+ records, improving data accuracy and enabling dependable weekly KPI tracking.
Automated Excel-based reporting workflows using advanced lookup and Pivot-tables, reducing reporting effort by 40%.
Built 2 Power BI dashboards to visualize revenue trends, customer mix, and monthly performance metrics.
Collaborated with cross-functional teams to define and validate 6 KPIs, improving reporting consistency by 15%.

My Projects

Credit Card Fraud Detection: Cost-Optimized ML Classifier

Tackled severe class imbalance (0.17% fraud) across 56,746 transactions. Benchmarked 4 models (XGBoost, Random Forest, LightGBM, Logistic Regression) with SMOTE balancing and SHAP explainability. XGBoost hit ROC-AUC 0.977 with a cost-tuned threshold — projecting $26.6M annual savings (78.9% cost reduction).

See My Work

Customer Lifetime Value (CLTV) Prediction

Built a probabilistic CLTV engine using BG/NBD and Gamma-Gamma models on e-commerce RFM data to predict future customer value. Segmented customers into priority tiers, enabling the business to focus retention marketing spend on the highest-value segments instead of blanket campaigns.

See My Work

Conversational RAG Agent for Business Intelligence

Built a Retrieval-Augmented Generation (RAG) agent with semantic indexing that lets stakeholders query business data in plain English and get instant, source-grounded answers — eliminating analyst turnaround delays. Evaluated response quality with RAGAS metrics.

See My Work