30-second tour: start with Fraud Detection → Healthcare Claims Intelligence → Loan Default Prediction. Those three show the full loop: data → model → API → dashboard → CI/CD.
Senior Data Analyst with 5+ years across healthcare, banking, and fintech. I build scalable SQL pipelines, predictive models, and executive dashboards, and I ship them as working products rather than slide decks. Every deliverable answers three questions: what happened, why it happened, and what to do next.
rahul = {
"role": "Senior Data Analyst · BI Developer",
"based_in": "New York, USA",
"domains": ["Healthcare", "Banking", "Fintech"],
"core_stack": ["SQL", "Python", "Power BI", "Tableau", "Snowflake"],
"education": "M.S. Data Analytics — Montclair State University (2025)",
"philosophy": "Every dataset is a narrative waiting to be told.",
"exploring": ["SHAP explainability", "drift monitoring", "dbt lineage"],
"open_to": ["Data Analyst", "Sr. Data Analyst", "BI Developer", "Analytics Engineer"],
}How I approach a problem
flowchart LR
Q["Business question"] --> D["Data audit"] --> E["Exploration"] --> M["Model / hypothesis"] --> S["Dashboard or API"] --> R["Stakeholder readout"]
| member records queried & modeled |
reporting effort via automation |
SQL query performance |
AUC on readmission- risk model (pilot) |
| Healthcare Utilization analytics · readmission risk · care-gap intelligence |
Banking Portfolio performance · credit risk · executive KPI scorecards |
Fintech Fraud detection · payment analytics · operational reporting |
| Languages & Analysis |
|
| Machine Learning |
|
| BI & Visualization |
|
| Databases |
|
| Apps, APIs & DevOps |
|
| Methods |
|
Ordered from most to least complex. Complexity reflects scope (modeling, deployment, testing, architecture), not how much I enjoyed a project.
Complexity ▰▰▰▰▰ · ML + API + dashboard + containerization + CI
Python scikit-learn SQLite Streamlit Docker GitHub Actions
A transaction-monitoring engine that blends deterministic rules with supervised ML and unsupervised anomaly detection, trained on the public European card-fraud benchmark dataset.
- Hybrid scoring: 60% ML fraud probability, 20% anomaly risk, 20% rule score, mapped to approve / review / block bands
- SQL-backed monitoring: fraud trends, success rates, channel-level risk, and a high-risk alert table
- Analyst tooling: Streamlit dashboard with a live transaction-scoring form
- Engineering: Dockerized, unit-tested, CI on every push
View scoring architecture
flowchart LR
T["Transaction"] --> R["Rule engine · 20%"]
T --> M["Fraud classifier · 60%"]
T --> A["Anomaly detector · 20%"]
R --> H{{"Hybrid risk score"}}
M --> H
A --> H
H -->|"below 45"| L["Approve"]
H -->|"45 to 75"| V["Review"]
H -->|"75 and above"| B["Block"]
Python FastAPI scikit-learn PostgreSQL Streamlit Docker GitHub Actions Render
A payer-analytics workflow that mirrors real care-management work: engineer the data, validate it, model readmission risk, publish dashboard-ready marts, and serve predictions from a deployed API.
- Data quality: automated validation suite (keys, foreign keys, date logic, financial logic), 10/10 checks passing
- Analytics marts: utilization trends, cost drivers, provider performance, and cohort analysis for BI tools
- Model: readmission classifier with ROC-AUC 0.74 on 18K synthetic claims
- Serving:
/predictand/predict/batchendpoints with Swagger docs; CI runs the pipeline, Ruff, and Pytest
▶ Live dashboard · API docs · </> Source
View pipeline architecture
flowchart LR
G["Generate claims + members"] --> V["Validate · 10 checks"]
V --> T["Train readmission model"]
V --> D["Build dashboard marts"]
T --> API["FastAPI · /predict"]
D --> BI["Power BI / Tableau / Streamlit"]
Python XGBoost scikit-learn FastAPI Streamlit
An end-to-end credit-risk system: feature-engineered loan applications, a gradient-boosted classifier with calibrated probabilities, a FastAPI scoring service, and a Streamlit dashboard built for risk officers.
Complexity ▰▰▰▰▱ · Modeling rigor, experimentation, and spatial reasoning
| Project | What it does | Stack | Links |
|---|---|---|---|
| Time-Series Forecasting Studio | Benchmarks ARIMA / SARIMA against ML regressors on sales and demand series, with seasonality decomposition, residual diagnostics, MAPE / RMSE comparison, and CSV-exportable forecasts | Statsmodels Prophet Scikit-learn Streamlit |
Live · Code |
| A/B Testing Analytics Dashboard | Experimentation workbench that accepts uploaded, simulated, or Kaggle data, runs frequentist tests, computes power and lift, and turns p-values into plain-English decisions | SciPy Statsmodels Plotly Streamlit |
Live · Code |
| Bioscope · Biodiversity Risk | Geospatial compliance assistant for New Jersey businesses: scores ecological risk by location from a species and habitat database, recommends mitigations, and renders interactive maps | Flask PostgreSQL Leaflet.js REST API Vercel |
Live · Code |
Complexity ▰▰▰▱▱ · Relational design, authentication, role-based access
| Project | What it does | Stack | Links |
|---|---|---|---|
| Tax Tracker System | Lets finance teams log company tax records, monitor pending payments, and audit activity, with role-based access modeled on real CFO workflows | Flask MySQL SQLAlchemy RBAC Render |
Live · Code |
| Student Management System | Academic operations platform covering registration, courses, grade entry, and performance analytics on a relational schema with secure auth | Flask SQLAlchemy MySQL Bootstrap |
Live · Code |
Complexity ▰▰▱▱▱ · Where the habit of asking good questions started
| Project | What it does | Stack | Links |
|---|---|---|---|
| Airbnb NYC · EDA Studio | Explores 39K+ listings end to end: pricing distributions, neighborhood and room-type demand, geographic heatmaps, and host concentration | Pandas Plotly Seaborn Streamlit |
Live · Code |
| IPL Data Visualization (2008–2017) | A decade of Indian Premier League data: toss impact on win rate, venue advantage, run-rate dynamics, and top performers, told through narrative visuals | Jupyter Pandas Matplotlib Seaborn |
Live · Code |
| Period | Role | Selected impact |
|---|---|---|
| Aug 2024 – Present | Senior Data Analyst Optum Health · New York |
SQL across Snowflake, Oracle, and SQL Server on 120M+ member records Cohort and utilization analyses that lifted member retention by 18% Readmission-risk models (AUC 0.87) that cut 30-day readmissions by 12% in pilot cohorts |
| May 2021 – Aug 2023 | Data Analyst Fiserv · Hyderabad |
Analyzed 2M+ daily transactions and lifted processing success rate by 3.4 points Cut insight latency from 24 hours to under 2 hours with Oracle and PostgreSQL ETL Fraud, merchant, and SLA dashboards used daily by 15+ stakeholders |
| Jan 2020 – May 2021 | Data Analyst Wells Fargo · Hyderabad |
Analyzed a $3B+ loan portfolio for origination, delinquency, and risk reviews Flagged $480K+ in suspect activity with SQL and Python anomaly screening Automated month-end reporting, closing the cycle 3 business days faster |
Education: M.S. in Data Analytics, Montclair State University (May 2025)
| Roles I'm looking for Data Analyst · Senior Data Analyst · BI Developer Analytics Engineer · Product / Decision Science |
How I like to work Full-time or contract · Hybrid or remote US-based, currently authorized to work in the US |
"Every dataset is a narrative waiting to be told."