From Spreadsheet Scars to Automated Systems: Why I Build Data Engines

1. The Midnight CSV Trap

I did not start my career in a computer science research lab or an enterprise cloud engineering team. My journey began in the messy trenches of commercial e-commerce operations at Vaithuhay, managing pre-order campaigns for high-ticket tech accessories with average order values reaching 2M to 10M+ VND.

My unglamorous reality back then was a familiar one: sitting at my desk at 11 PM, manually exporting 15 different CSV reports from Meta Ads Manager, Shopee, and internal order sheets, then desperately stitching them together with fragile VLOOKUPs and nested IF statements. One misplaced column index or shifted row could silently corrupt a customer acquisition cost (CAC) calculation, risking hundreds of millions of VND in marketing budget misallocation.

That was the moment I made a promise to myself: I will never accept manual spreadsheet toil as a sustainable way of working. Reliable automated systems are the only answer.

2. Consolidating 100B+ VND for the CFO & The Lesson of Data Truth

When I joined SAGEN GROUPE as a Digital Marketing & E-Commerce Data Analyst, the stakes escalated dramatically. I was tasked with modeling multi-brand financial P&L across more than 12 healthcare and cosmetic brands (such as Arkopharma, Phyto, and Embryolisse Vietnam), representing over 100B+ VND in annual corporate scope.

The existing workflow was consuming 2 to 3 hours every single morning just to copy, paste, and reconcile yesterday's transactions. I refused to let our team burn cognitive energy on robotic tasks. I mastered Power Query (M-Code) to build dynamic, self-refreshing financial ETL transformations, and engineered the company's first automated Google Apps Script API ingestion pipelines connecting Shopee Open Platform, TikTok Shop, Haravan ERP, and Meta Graph API directly into Looker Studio BI.

Working directly alongside the CFO taught me a fundamental lesson that no textbook can teach: data modeling is not about showing off convoluted dashboards. It is about mathematical truth:

If a pipeline shows rosy numbers but cash flow is leaking, the model is broken. Ground truth always wins.

3. Transparent About AI: A Force Multiplier, Not a Disguise

In today's tech market, many candidates try to present themselves as 10-year veteran fullstack software engineers who hand-coded every distributed database from scratch. I believe in radical honesty: I heavily leverage AI coding partners (Claude Code, Hermes Agent, Model Context Protocol MCP) as high-velocity technical collaborators.

When I architected the P01 Omnichannel Delivery WebApp using Next.js 16 and client-side SQLite WASM, or deployed containerized microservices on GCP Cloud Run, AI acted as an elite co-pilot. It allowed me to translate relational business architecture into working, battle-tested production code in days rather than months.

However, AI without deep domain understanding is a recipe for hallucinations. AI does not know why the Shopee API timestamps skew, how courier name variations break 63-province Vietnamese geocoding, or why 15-day HMAC-SHA256 chunking is required to prevent gateway timeouts. I bring the commercial context, relational schemas, and strict data validation; AI provides the acceleration.

4. The Discipline of the Craft

AI tools come and go, but core engineering fundamentals remain durable. Because I respect the craft, I deliberately invest hours into rigorous, unassisted self-development:

5. The Road to Data Engineering & What I Bring to Your Team

Today, as an independent consultant for Instax Vietnam (FUJIFILM) (>3.5B+ monthly GMV) and SAGEN GROUPE, I design 24/7 serverless analytics portals, model full-funnel attribution (demonstrating that Video Reels deliver 1.9x efficiency over static formats), and build zero-cost data delivery pipelines.

My trajectory is clearly set on evolving from Commercial Data Analytics into Data Engineering (DE)—architecting resilient API ingestion pipelines, Medallion Lakehouses, and automated data pipelines that transform messy operational chaos into pristine, decision-ready data marts.

Here is what I bring to a high-velocity data team on Day 1:

• Business & P&L Empathy: I understand contribution margin, CAC, ROAS, and inventory velocity natively. I build data systems to drive revenue and profit, not just to write code.
• Zero Tolerance for Manual Toil: I proactively automate repetitive data extraction and cleansing before it accumulates into technical debt and wasted work hours.
• High Agency & Relentless Learnability: I embrace new tools rapidly, accept technical critique with humility, and take full ownership of problems from end to end.

If you are building data systems that matter and value authenticity, relentless drive, and real commercial impact—let's connect.