DATA ANALYTICS ENGINEER

Thai Trung Kien

I build tested data platforms that people actually trust.

Pharmacist turned analytics engineer — nearly 3 years building data platforms for healthcare, retail and community products. I've modeled 843M+ rows into tested dbt warehouses, cut data latency from daily batch to under 5 minutes with Debezium/Kafka CDC, and ship dashboards trusted by ~900 business users — every number backed by 150+ automated tests, not hope.

View case studies → Contact me
GitHub in Email

kienthai2711@gmail.com · +84 945 613 721
HCM City · Hanoi, Vietnam

843M+

rows processed in one project

150+

automated data & unit tests shipped

900

business users served at VMT

100K

members in Xóm Data community

// 00 — WORK EXPERIENCE

A journey of impact through data

Data Analytics Engineer

VMT Holdings · Nov 2025 – Present
  • Built VDP — VMT's retail data platform — end to end: 4 OLTP sources (Oracle ERP, Pancake social commerce, internal systems) through staging → warehouse → marts to a governed semantic layer serving Tableau and Metabase.
  • Debezium CDC + Kafka + Spark on Kubernetes — data latency from daily batch to under 5 minutes.
  • dbt models with schema & data-quality tests; orchestration via Airflow, Node.js services and n8n.
  • Operational BI suite serving nearly 900 business users across departments.

Analytics Engineer

Mevis Healthcare · Sep 2025 – Nov 2025
  • Refactored core ETL to support high-concurrency access for the Sales/Operations portal, ensuring 99% data availability for business users.
  • Engineered the underlying data infrastructure for a high-concurrency user portal, enabling real-time data access for thousands of users.
  • Deployed Looker Studio dashboards converting raw operational logs into decision-ready insights for leadership.

Data Analyst

Phuong Dong General Hospital & Asahi · Jan 2025 – Jul 2025
  • Centralized BigQuery warehouse integrating CRM and HIS (Hospital Information System) data, automating daily reporting workflows.
  • Omnichannel funnel analysis with Marketing & Sales — conversion trends that drove a 17% increase in marketing efficiency.
  • Sales & Revenue dashboards (traffic, conversion, revenue per patient) for the Board of Directors; department-level performance analysis cut operational costs by 10%.
  • Automated YouTube analytics extraction with n8n, feeding Google Sheets and Looker Studio for holistic platform reporting.

Moderator & Admin

Xóm Data Community · Apr 2025 – Present
  • Scaled Vietnam's leading data community from 0 to 100,000+ active members in 14 months.
  • Run a free self-hosted SQL Server practice environment (14 schemas) and the Xóm Jobs connect series bridging data professionals and employers.
  • Contributed to the Xóm Dataset web platform and conducted CV & portfolio reviews for members; enforce strict moderation to keep the community spam-free and hands-on.

BI Developer & Excel Trainer

ComeEco Company · May 2025 – Jul 2025
  • Designed a Financial Reporting System in Power BI with complex DAX measures for YoY and MoM growth.
  • Automated workflows delivering real-time financial forecasts to the CEO — replaced manual Excel reports, saving 50% of reporting time.
  • Ran internal “Data Thinking & Visualization” training to foster a data-driven culture.

// 01 — FEATURED CASE STUDIES

Data platforms, end to end

Every build follows the same arc: a real decision problem → a modeled, tested pipeline → a measurable outcome someone can open and question.

CS-01 · AVIATION CX ANALYTICS

AeroVoice Intelligence Platform

Shipped an auditable analytics platform that turns 214,681 aviation reviews into governed, decision-ready insight — 4 conformed review grains, semantic marts, local NLP, and a privacy-safe public dashboard that answered 15 stakeholder requests with zero raw text exposed.

39 dbt models 95 dbt tests 15 stakeholder requests

SQL Server · dbt Core · Airflow · Python NLP · ECharts

AeroVoice dashboard overview

CS-02 · MARKET-ENTRY DECISION PRODUCT

Green SM NYC Mobility Intelligence

Reduced a $-scale market-entry question to two testable corridors by processing 843.7M official NYC trip records through a tested dbt dimensional model (61/61 nodes passing) — explicitly separating observed activity from unmet demand.

843.7M rows 263 zones 36 months

Parquet · dbt · Kimball modeling · JS dashboard · GitHub Pages CI/CD

CS-03 · BATCH & CDC PIPELINE

VeloCiSync M2MR

Guaranteed zero-loss, zero-duplicate delivery across a dual-path pipeline — Spark backfills through MySQL staging plus Debezium/Kafka CDC into MongoDB and Redis, where offsets commit only after both targets succeed. 438K events validated end to end.

438K events 46.33 rec/s CDC 14 tests + CI

Python · Spark · Debezium · Kafka · MongoDB · Redis · Docker

MySQL → Debezium → Kafka

↓ checkpoint-safe consumer

MongoDB + Redis projections

JSONL backfill → Spark → staging merge

CS-04 · DATA + AI PLATFORM

Polaris Data Jobs Platform

Replaced keyword job-hunting with semantic matching — an automated platform unifying postings from 3 sources into a traceable PostgreSQL model, ranking CV-to-job fit with pgvector embeddings, served through a Next.js portal on a 9-container stack with 21 automated tests.

3 job sources 384-dim embeddings 21 automated tests

Python · Dagster · dbt · PostgreSQL + pgvector · Next.js · MinIO

TopCV · LinkedIn · ITViec

↓ scrape → raw HTML in MinIO

↓ PostgreSQL + pgvector · dbt models

↓ 384-dim embeddings · cosine match

Next.js portal · Metabase · job alerts

CS-05 · EVENT-DRIVEN STREAMING PLATFORM

AtmosFlow — Real-time Weather Platform

Turned a raw weather API into analytics-ready city summaries — an event-driven Kafka pipeline streaming live weather for 4 Vietnamese cities into PostgreSQL, refined through a dbt Bronze–Silver–Gold workflow with full lineage, refreshed every 15 minutes by Airflow and monitored in Grafana.

Bronze–Silver–Gold 15-min refresh Grafana observability

Python · Kafka · PostgreSQL · dbt · Airflow · Grafana · Docker

Weather API · 4 cities

↓ producer → Kafka stream

↓ Bronze → Silver → Gold (dbt)

↓ Airflow every 15 min

Grafana watches pipeline + freshness

CS-06 · SERVERLESS MEDALLION ELT

Gold Price Platform — VN Market Tracker

Automated a daily market-intelligence loop with zero servers — GitHub Actions ingests Vietnam gold prices for 8 brands into immutable Bronze snapshots on GCS, dbt on BigQuery models Silver/Gold trends, spreads and day-over-day volatility, and a self-updating README dashboard publishes the result with Discord alerts.

Medallion ELT 100% unattended $0 infra

Python · GitHub Actions · dbt · GCS + BigQuery · Discord webhooks

8 VN gold brands · daily cron

↓ Bronze snapshots on GCS

↓ dbt on BigQuery → trends · spreads

↓ self-updating README charts

Discord alert on price movement

CS-07 · SOFTWARE-GRADE ANALYTICS AUTOMATION

XSMB Analytics — Lottery Data Pipeline

Collapsed a manual collect-analyze-report cycle into one CLI command — a typed, modular Python pipeline where pydantic validates every draw at the boundary, incremental CSV history stays append-only and deduplicated, and a Plotly dashboard of frequency/recency statistics auto-publishes to GitHub Pages daily.

30+ tests strict mypy daily CI/CD

Python · pydantic · Typer · Plotly · GitHub Actions · GitHub Pages

Daily XSMB draws

↓ pydantic typed validation

↓ incremental CSV · dedup

↓ frequency · recency analytics

Plotly dashboard → GitHub Pages

// 02 — LIVE DASHBOARDS

See the outputs, not just the code

Every platform ends in something a stakeholder can open. These are deployed and public — click through and poke at the filters.

AeroVoice overview dashboard

AeroVoice — Airline rankings

ECharts · GitHub Pages

↗
AeroVoice NLP sentiment panel

AeroVoice — NLP sentiment & topics

VADER + NMF · privacy-safe aggregates

↗

Green SM — NYC mobility decision views

JS dashboard · GitHub Pages

↗
AeroVoice coverage dashboard

AeroVoice — Coverage & quality gates

sample-size guardrails · governed metrics

↗

// 03 — MORE PROJECTS

Selected work from the archive

{{ p.cat }}

{{ p.title }}

{{ p.desc }}

KEY RESULT — {{ p.result }}

// 04 — SKILLS

One full platform loop

HEALTHCARE & PHARMA

Revenue cycle, patient flow, HIS/CRM integration — grounded in a pharmacy degree from HUP

SALES & MARKETING

RFM segmentation, retention, conversion funnels, ROI/ROAS, omnichannel attribution

RETAIL & E-COMMERCE OPS

VMT Holding's retail data platform: Oracle ERP + Pancake social commerce, sales & inventory operations, customer segmentation & retention, finance/debt tracking for ~900 business users

BANKING & HR

Credit/deposit/payment-account segmentation, risk profiling, workforce attrition analytics

DATA STORYTELLING

Translating models and metrics into decisions for non-technical stakeholders and boards

LEADERSHIP & COMMUNITY

Moderator & Admin of Xóm Data (100K+ members); mentoring, CV reviews, job connect

RIGOR & LEARNING

High accuracy with sensitive healthcare data; self-taught Airflow, Docker, dbt, Kafka

INGEST & COLLECT

Kafka Debezium CDC Spark Python Selenium BeautifulSoup REST APIs

STORE & MANAGE

BigQuery PostgreSQL SQL Server MySQL Oracle MongoDB Redis MinIO

TRANSFORM & MODEL

dbt Core Spark Advanced SQL Kimball star schema Data Vault 2.0 Pandas

ORCHESTRATE

Airflow Dagster n8n

DATA QUALITY

dbt tests Data contracts Audit reconciliation Replay-safe design Python unit & integration tests

GOVERNANCE & CATALOG

dbt Docs Governed metric definitions Operational runbooks

BI & ANALYTICS

Power BI Tableau Looker Studio Metabase Cube ECharts GA4

CI/CD & GITOPS

Git & GitHub Actions Docker Kubernetes Makefile

PROJECT & TOOLS

Jira Notion Trello Figma VS Code DBeaver Jupyter

01 / INGESTION & STREAMING

Kafka Spark Python Debezium CDC REST APIs

02 / STORAGE & WAREHOUSING

BigQuery PostgreSQL SQL Server MySQL MongoDB Redis MinIO

03 / TRANSFORMATION & MODELING

dbt Core Pandas Kimball star schema Data Vault 2.0 Advanced SQL

04 / QUALITY & TESTING

dbt data tests Data contracts Audit reconciliation Replay-safe design

05 / ORCHESTRATION & AUTOMATION

Airflow n8n GitHub Actions Docker Dagster

06 / SEMANTIC LAYER & BI

Power BI Tableau Looker Studio Cube Metabase ECharts

// 04b — HOW MY DATA FLOWS

Source to decision, one governed path

SOURCES Oracle ERP orders · finance Pancake social commerce Internal OLTP MySQL · Postgres · MSSQL Web & APIs scraping · RESTINGEST & COLLECT Debezium → Kafka CDC · <5 min Checkpoint consume commit after OK Spark backfills replay-safe batch Python scrapers Selenium · BSoup Dagster jobs assets · Redis locksSTAGING & LAKE MinIO raw · immutable Staging schemas idempotent · dedup DuckDB Parquet · localWAREHOUSE PostgreSQL operational marts BigQuery cloud analytics SQL Server on-prem martsTRANSFORM dbt models stg → int → marts Spark jobs heavy aggregations Semantic marts conformed dimsDATA QUALITY dbt tests schema · values Data contracts validate · quarantine Audit accounting reconciliationGOVERNANCE dbt docs lineage · catalog Governed metrics one definition Runbooks ops · monitoringSEMANTIC LAYER Cube metrics API Metric definitions shared logicBI & ACTIVATION Tableau · Power BI ~900 users Metabase · Looker ECharts dashboards Mongo · Redis low-latency serving n8n alerts reports · notify

ORCHESTRATION LAYER

Airflow · Dagster · n8n

schedules ingestion dbt runs data-quality checks alert & report jobs

INFRASTRUCTURE & CI/CD

Docker · Kubernetes · GitHub Actions

Docker Compose stacks Spark on K8s PR quality gates runbooks & monitoring

solid = batch flow · dashed = CDC streaming · dotted = automation · the two layers above schedule and run everything

// 05 — ABOUT

From pharmacy to pipelines

I trained in pharmacy at Hanoi University of Pharmacy and worked through nearly every role the industry offered — sales, medical rep, marketing, medical affairs. I could do the work, but nothing pulled me back to the laptop after hours. The turn came in medical affairs, when a mentor pointed at a chart in a research paper and asked: "Work out how they computed this." I spent days re-deriving every number — and realized what hooked me wasn't pharma, it was the mechanism behind a number.

So I moved into data — one real problem at a time: Python when Excel stopped scaling, SQL when I got tired of waiting for extracts, Power BI when analyses needed an audience, then dbt, warehousing and pipelines when I saw the hard part was never the dashboard but everything underneath it. Never a tool because it was hot — only because a problem demanded it.

Today, as an analytics engineer, my job in one sentence: help the business trust its data before using it to decide. Pharmacy's discipline carried over intact — ask how a number was made before accepting it, because one wrong assumption changes every conclusion downstream.

SKILL DEPTH ACROSS THE JOURNEY · PHARMACY → DATA ANALYST → ANALYTICS ENGINEER

PHARMA

MARKETING

DATA ANALYST

AE (NOW)

{{ row.skill }}

brighter = deeper practice at that stage

Don't just learn how to find the answer — learn to understand why that answer exists.

Outside work I help run Xóm Data, Vietnam's data community of 100,000+ members, and teach — because explaining something is the fastest way to find out what you don't actually understand yet.

CURRENTLY

Data Analytics Engineer @ VMT Holdings — CDC streaming + dbt warehouse serving ~900 users

COMMUNITY

Moderator & Admin, Xóm Data — 100K+ members, 14-schema free SQL practice server

CERTIFICATIONS

Google Data Analytics, Advanced DA & BI Certificates; HackerRank SQL (Advanced)

DOMAIN EXPERTISE

Healthcare, Pharma, Marketing and Retail — from hospital data to omnichannel commerce

TECHNICAL STRENGTH

Full loop: ingestion → storage → modeling → testing → semantic layer → dashboards

2026

Ready for new challenges as a Middle Analytics Engineer

// 06 — EDUCATION & CERTIFICATIONS

Hanoi University of Pharmacy

Faculty of Pharmaceutical Management and Economics — systems thinking and process management (2019 – 2024).

Google Data Analytics Professional

COURSERA · 2024

Google Advanced Data Analytics

COURSERA · 2024

Google Business Intelligence

COURSERA · 2024

SQL (Advanced) Certificate

HACKERRANK · 2024