Data Analyst · Business Intelligence · Reporting

Thrinesh VuribindiI turn complex data into clear insights that help teams understand what’s happening, why it matters, and where to act.

For 5 years I have worked across healthcare and financial services data. I frame the business question, clean and model the data in SQL and Python, build the dashboards in Power BI and Tableau, and explain the result in plain language to the people making the call.

  • 5 years experience
  • Healthcare and financial services
  • Microsoft Certified: Fabric Analytics Engineer (DP-600)
  • Chicago, IL
See my work LinkedIn GitHub
Portrait of Thrinesh Vuribindi
Experience

Each role added a layer.

Today I work at enterprise scale, where the data spans many platforms and the reporting has to hold up to audits. Before that, I built BI reporting that turned stakeholder requests into dashboards people trust. I started in financial analysis, finding the cost drivers in the numbers.

2025May 2025 to present

Analytics at enterprise scale

Data Analyst, UnitedHealth Group · Healthcare

Now the work spans many systems and business units, so my focus is reliable pipelines, traceable numbers, and reporting that people can serve themselves.

  • 12M+ recordsAnalyze monthly claims and eligibility data in Snowflake with SQL and SAS. Traced lineage across Snowflake, SQL Server, Oracle, and SAP HANA to resolve 19 claims discrepancies before CMS submission.
  • 3 days → <1 dayAutomated 7 recurring reporting workflows with Python, SSIS, and Fabric pipelines, saving about 12 hours of manual work a week.
  • 40+ KPIsDelivered 6 production Power BI dashboards across utilization, reimbursement, quality, revenue cycle, and cost for 15+ stakeholders in 8 business units, and trained 30 users across 3 teams to self-serve them.
  • 10+ sourcesIntegrated source systems through Azure Data Factory and ADLS Gen2 and standardized 33 fact and dimension tables in Microsoft Fabric with Data Engineering.
2022Jan 2022 to Aug 2023

Turning requests into reporting people trust

Data Analyst, Infosys

I learned that a dashboard is only as useful as the agreement behind its metrics, so I started every build with requirements, definitions, and validation.

  • 20+ KPIsDefined KPIs and dashboard acceptance criteria through 10 requirements sessions with 12 stakeholders, closing 6 reporting gaps.
  • 2 hrs → 20 minConsolidated 4 Excel, Google Sheets, and SSRS reports into one Power BI solution with star-schema models, serving 150+ users.
  • 90% automatedAutomated pre-refresh validation on HIPAA-regulated data that had been fully manual, with zero discrepancies reaching users.
2020Jun 2020 to Dec 2021

Finding the numbers that matter

Associate Data Analyst, Genpact · Financial services client

Variance analysis and forecasting in Advanced Excel taught me to start from the business question and work back to the data.

  • $180K / 8%Contributed to annual operating-cost savings by analyzing revenue, cost, gross margin, and backlog variance and presenting cost drivers to business owners.
  • 13 categoriesBuilt forecasting and budgeting models with Power Pivot, PivotTables, and SUMIFS, and automated actual-to-budget comparisons for monthly planning.
  • 4 hrs → 30 minCut weekly report preparation by rebuilding the manual steps in Power Query.
Projects

Three problems, worked end to end.

Each project starts with a business question, shows the method, and ends with a recommendation and its limits. All code, notebooks, and reports are on GitHub.

Operations analytics · Data pipeline · BI

Surgical Scheduling Variance Analysis

How accurately does a hospital estimate surgery length, what do the misses cost, and which procedures should be fixed first?

  • DataLoaded about 1.88M raw rows from a de-identified UC Irvine surgical dataset into a Microsoft Fabric lakehouse with bronze, silver, and gold layers.
  • QualityRemoved 1,366 duplicates, repaired 13 broken timestamps, flagged every change, and documented each row count from 65,728 records to a 48,118-case cohort.
  • MethodBenchmarked each of 418 procedures on its historical median, since a few very long surgeries skew the mean. Tested the minimum-volume rule before fixing it at 30 cases.
  • DecisionRecommended reviewing 133 procedures first, cleaning up 207 procedure names that mix different operations, and starting to record booked durations.
Microsoft FabricPySparkSpark SQLPower BIDAX (17 measures)Data quality
Open live dashboard ↗ Code and docs ↗
$15.8M to $27.1Mestimated yearly cost exposure from scheduling variance
53.3%of cases missed their benchmark by more than 30 minutes
133 of 418procedures drive 53.6% of the variation
2,167 hrsof room time lost each year to early finishes alone

Data cleaning checkpoints, rows kept

Raw records65,728 Duplicates removed64,362 Usable duration57,861 Cases in scope48,118

Cost range uses illustrative rates of $35 to $60 per OR minute and is a planning scenario, not a measured loss.

Payment integrity · SQL modeling · Cost analysis

Medicare Claims Payment Integrity Analytics

Every claim went to a manual reviewer, yet bad claims still got paid. Which claims can be decided safely by rules, and when does automation stop saving money?

  • DataCleaned 10 messy extracts in Python (mixed date formats, overlapping files, duplicate remittance batches, NPI check-digit validation) and loaded staging, core, and mart schemas in PostgreSQL.
  • MethodWrote 10 SQL prepayment edits from published Medicare billing rules and scored them against the payer's final decision with a confusion matrix.
  • FindingDenials rose from 5.8% to 6.8% in 2025 to 8.2% in 2026. The 296 missed claims were mostly medical necessity and missing information, which need records review, not more rules.
  • DecisionRecommended 6 edits that auto-deny, 4 that route to a reviewer, and provider outreach starting with the top quartile, which accounts for 63% of flagged claims.
SQLPostgreSQLPython (pandas)Power BIWhat-if analysisSensitivity analysis
Code and report ↗
$425K → $120Kmodeled decision cost, full review vs. rule-based routing
96%of 23,631 claims decided without manual review
82%of improper claims caught before payment
93.3%accuracy on the 1,485 claims the edits flagged

Manual reviews needed

Review everything23,631 Rule-based routing854 Break-even: rules stay cheaper until a missed claim costs over $1,281.

Synthetic claims built on real CMS file layouts and code sets. Dollar figures are a scenario with adjustable inputs in the report.

Statistical analysis · Time series

Statistical Time-Series Analysis

Do four common beliefs about stock behavior hold up when tested on five years of daily Apple market data?

  • DataCollected five years of daily Apple stock data (1,254 trading days), checked it for missing values and duplicates, and added calculated fields such as daily percentage change and 30-day volatility.
  • MethodTested each belief with t-tests, correlation analysis, a normality test, and monthly and weekday trend analysis.
  • FindingPrice and volume showed a -0.40 correlation, but daily returns and volume showed almost none (0.01). The first number came from two opposite long-term trends, not a real relationship.
  • TakeawayHow a metric is defined can change the conclusion. Volatility in dollars looked like rising risk, while volatility in percentages showed it stayed stable.
PythonpandasSciPyHypothesis testingCorrelation analysisMatplotlibSeaborn
Notebook and findings ↗

Four beliefs, tested

Higher price means higher risksteady at 1.65% dailyNot supported
Volume predicts price directionp = 0.267Not supported
Calendar patterns repeatchanged year to yearInconclusive
Daily returns are normalkurtosis 6.64Not supported
1,254trading days analyzed, Aug 2021 to Aug 2026
3 of 4common assumptions not supported by the data
Capabilities

What I bring to a team.

Grouped by the work, not the tool. Each area reflects work I have done across my roles and projects.

Problem framing and requirements

Turning a vague ask into a defined question, agreed KPIs, and acceptance criteria before building anything.

Requirements gathering · KPI definition · Acceptance criteria · UAT · Agile · JIRA

Data preparation and quality

Profiling, cleaning, reconciling, and tracing data so the numbers hold up when someone asks where they came from.

Data profiling · Validation · Reconciliation · Data lineage · Source-to-target mapping · Data dictionaries

Analysis and statistics

Choosing the right measure and test, then saying clearly what the evidence does and does not support.

Hypothesis testing · A/B testing · Correlation · Trend and seasonality · Forecasting · Variance and root-cause analysis

BI and reporting

Star-schema models, tested DAX, and dashboards built so stakeholders can answer their own follow-up questions.

Power BI (DAX, Power Query, RLS, Service) · Tableau · SSRS · Advanced Excel · Microsoft Copilot

Query, code, and platforms

Writing the SQL and Python that move data from source systems into models people can use.

SQL (CTEs, window functions, T-SQL) · Python · SAS · PySpark · Snowflake · Microsoft Fabric · Databricks · SQL Server · Oracle · PostgreSQL · Azure Data Factory · SSIS · dbt

Communication and enablement

Explaining findings in plain language, stating limits up front, and documenting so others can work without me.

Stakeholder presentations · User training · Documentation · Cross-team collaboration with IT and Data Engineering
Credentials

Certification and education

Microsoft Certified

Fabric Analytics Engineer Associate (DP-600)

Verify credential ↗

2023 to 2025

M.S. Information Systems

Saint Louis University, St. Louis, MO

2017 to 2021

B.Tech, Electrical and Electronics Engineering

Guru Nanak Institutions Technical Campus, Hyderabad, India

Contact

Have a question your data should answer?

I am open to Data Analyst, BI Analyst, and Reporting Analyst roles. Email me and I will send a resume matched to your role.

Emailvuribindithrinesh@gmail.com
Phone+1 (312) 283-9011
LinkedInlinkedin.com/in/thrineshv
GitHubgithub.com/thrinesh13