Data Analysis Projects

Data quality checker

Python Pandas CSV

Context: Reporting work often starts from messy exports. Before those files feed a dashboard or a SQL load, they need a pass/fail check for schema, nulls, and duplicate rows.

What I did: Wrote a single Pandas script that loads a sample orders CSV as text, then runs a contract of rules: expected columns and types, required fields, full-row duplicates, duplicate order IDs, quantity greater than zero, and allowed status/product values. Failed checks print Excel-style row numbers so the source file can be opened and fixed.

Outcome: Against a 10-row sample with planted issues, the checker passed column names and failed the seven remaining rules (types, nulls, duplicates, range, and allowed values). A non-zero exit code marks the run as failed for a scheduled job.

Source Script

[PASS] Schema - column names
[FAIL] Schema - value types
[FAIL] Nulls - required fields
[FAIL] Duplicates - full rows
[FAIL] Duplicates - business key
[FAIL] Range - quantity > 0
[FAIL] Allowed values - status
[FAIL] Allowed values - product

SUMMARY: 1 passed, 7 failed

Terminal report from the sample file: planted issues in dates, prices, blank customer IDs, duplicate order 1003, negative quantity, and typos like SHIPPD.

EXPECTED_COLUMNS = {
    "order_id": "int",
    "customer_id": "string",
    "order_date": "datetime",
    "product": "string",
    "quantity": "int",
    "unit_price": "float",
    "status": "string",
}

The contract at the top of the script: later checks read this instead of scattering rules through if-statements.

Tableau dashboards

Tableau Sample / training data

Context: Practice and portfolio dashboards built in Tableau (including Superstore-style sample data where noted).

What I did: Designed views for sales, marketing, and HR overview metrics — charts, filters, and layout for stakeholder scanning.

Outcome: Static screenshots on this page; interactive embeds are optional via Tableau Public if published later.

Python

Exploratory analysis and Excel automation with Pandas and NumPy. The data quality checker is written up as its own case study above. Highlighted notebooks below; more scripts live in the Python folder.

• Stock Market Case Study — Jupyter notebook analyzing market data (also available as rendered HTML).
• YouTube Data Analysis — case study notebook on channel / video metrics ( rendered HTML).
• Excel join comparison report — loads multiple workbooks, merges on shared keys with Pandas, and writes a reconciled Excel file.

# Merge security-report users with employee titles, then match estimated review items
id_match = pd.merge(user_name, titles_names, on="User ID", how="left")
no_pos_num = pd.merge(id_match, estimated_items, on="Position Number", how="left")
no_pos_num.to_excel("Authorized Infinium Users2.xlsx")

From the Excel comparison script: left joins across three sheets, then export.

SQL

Production-style reporting in SQL Server Management Studio (SSMS): multi-table joins, CASE logic, and filters for HR / ops exports. Sample queries are in the SQL folder.

• Ops org chart query — employee master + position + supervisor joins for a reporting hierarchy.
• Report Builder query (v2) — larger export with role mapping, payroll groups, and nested lookups.
• EEO open location list and other department-specific report extracts.

SELECT DISTINCT
  RIGHT(PRPMS.PREN, 6) AS EmployeeID,
  PRPMS.PRCKNM AS Name,
  PRPMS.PRL03 AS Dept,
  PRPMS.PRTITL AS Title,
  PRPMS_1.PRCKNM AS Reports_to
FROM psa.dbo.SMFD35_HRDBFA_PRPMS PRPMS (NOLOCK)
LEFT JOIN psa.dbo.SMFD35_HRDBFA_PEPOG PEPOG
  ON PRPMS.PRER = PEPOG.OGER AND PRPMS.PRPOS = PEPOG.OGPOS
LEFT JOIN psa.dbo.SMFD35_HRDBFA_PRPSP PRPSP (NOLOCK)
  ON PRPMS.PRER = PRPSP.SPER AND PRPMS.PREN = PRPSP.SPEN
LEFT JOIN psa.dbo.SMFD35_HRDBFA_PRPMS PRPMS_1 (NOLOCK)
  ON PRPSP.SPSPER = PRPMS_1.PRER AND PRPSP.SPSPEN = PRPMS_1.PREN
WHERE PRPMS.PRL01 = '001'
  AND PRPMS.PRL02 = 'XTT'
  AND PRPMS.PRTEDH = '0'
  AND PRPMS.PRSEC <> 'DRV'
ORDER BY PRPMS.PRCKNM ASC;

Excerpt from the ops org chart query: supervisor self-join pattern used for hierarchy reports.

Power BI

Desktop dashboard screenshot (interactive embed needs Power BI service publish-to-web).