Turning systems into scale

Tech Projects Workbench

This is where I prototype internal tools, governance workflows, and analytics that tie support, sales, and operations together. Not production apps. Working models that solve a real problem, clarify an idea, and show how better systems create better outcomes.

Each project started as a question: what if we could see this sooner, automate that step, or enforce this rule automatically? Then I built just enough to test the answer.

Case study

Olist Analytics Warehouse

Complete

A production-style analytics project that turns the Olist e-commerce dataset into a star-schema warehouse and a ten-notebook analytics suite. It covers the whole lifecycle, from raw Kaggle ingest to the story you’d tell a leadership team.

  • Ingested raw Olist data from Kaggle by API, staged it in notebooks, and modeled a fact and dimension star schema for customers, orders, payments, logistics, products, sellers, and geography.
  • Designed a reusable ETL pattern: documented CTEs, a table of contents per notebook, and separate bronze, silver, and gold transformations.
  • Built ten domain notebooks, each with narrative, SQL, and charts: order value, repeat rate, funnels, shipping delays, NPS against delivery time, seller reliability, and more.
DatabricksPythonSQLDuckDBPlotly

Prototypes

Guardrail: process governance for AI

Prototype

A compliance firewall for AI workflows. Guardrail proxies model calls, inspects responses against written policies, and decides whether to allow, redact, block, or escalate before content reaches a person.

YAML policiesAudit loggingLLM adapter

CRM Dupe Remover: data hygiene script

Prototype

A Python prototype that uses fuzzy matching and domain logic to find and merge duplicate CRM records before they reach campaigns, dashboards, or sales workflows. It became my reference model for large-scale deduplication in Zoho CRM.

PythonFuzzy matchingZoho CRM

Service Desk KPI Dashboard

Prototype

A dashboard that turns ticket exports into a control panel for support operations: SLA performance, resolution time, and volume patterns, so leaders see issues before they spike. It moved the team from gut feel to objective metrics.

Power BIStreamlitExcel exports

SQL Insights: trend detection sandbox

Prototype

A SQL playground with queries and mock datasets for exploring how support tickets cluster by product, region, and root cause, and how those patterns drive better investigations. A template for reporting layers on top of production data.

SQL ServerRetail datasetGitHub Pages