Portfolio 2026 · Tran Duong Thanh Phong

THEHYBRIDARCHITECT

Data Analyst · System Architect · Applied AI Researcher

CORE THESIS & METHODOLOGY

Analytics without adoption is overhead.Intuition without mathematical modeling is guesswork.

I bridge empirical academic research (MILP optimization, NLP, Hybrid Recommenders) with production cloud engineering (GCP, BigQuery, Multi-Agent Systems). The core philosophy: build resilient, self-sustaining data products that eliminate operational friction and deliver verified enterprise ROI.

01Problem Space

UNTANGLE THE MESS

Transforming 2.6M+ industrial IoT telemetry records, 6.7K unstructured customer reviews, and fragmented plant shift handovers into structured, solvable domain problems.

Execution Phase
02Engineering & Rigor

ARCHITECT THE ENGINE

Formulating Mixed-Integer Linear Programming (MILP space optimization), hybrid recommendation algorithms, and decentralized Google ADK multi-agent pipelines.

Execution Phase
03Business Impacts

BUILD SOLUTIONS THAT DRIVE BUSINESS IMPACTS

Deploying direct-to-chat automated shift reporting, zero-token BigQuery streaming pipelines, and 30–50% time-to-insight reductions that run reliably without manual overhead.

Execution Phase
Tran Duong Thanh Phong (Harris)
Ho Chi Minh City, VNHarris
ABOUT ME & PROFESSIONAL SUMMARY

TRAN DUONG THANH PHONG

Data Analyst · System Architect · Applied Researcher

Data Analyst with nearly 2 years of experience across manufacturing analytics, cloud-based data pipelines (GCP), and workflow automation. Proficient in SQL, Python, and modern BI platforms, with demonstrated success in translating complex operational requirements into actionable insights, interactive tracking dashboards, and AI-enabled digital solutions that eliminate manual overhead and accelerate data-driven decision-making.

~2 YOE
Enterprise & Mfg
4 Papers
Peer-Reviewed
E-Commerce
GPA 8.54/10 · UEL
CORE IDENTITIES

ONE LIFECYCLE. THREE ROLES.

01

THE BUILDER

Architecting cloud data pipelines, AI agents (GCP, BigQuery, Vertex AI), and headless workflow automation that eliminate operational friction.

GCPBigQueryVertex AIGoogle ADKPythonFastAPI
02

THE SCIENTIST

Formulating mathematical optimization models (MILP) and NLP research pipelines published in peer-reviewed academic conferences and journals.

MILP (PuLP)fastTextLDA Topic ModelingStatistical Modeling
03

THE ENABLER

Coordinating sprint backlogs, mentoring team members on pipelines, and leading plant-wide Data Literacy programs to foster self-service analytics.

Power BIData GovernanceSprint CoordinationData Literacy
TRACK RECORD & ROLES

WORK EXPERIENCE

01 / 03

Avery Dennison Vietnam

Data Analyst · Sep 2025 – Present

01
5+

Analytics Pipelines

40+

Team Members Trained

30–50%

Time-to-Insight Reduction

AI & Agentic Systems Engineering

Architected and developed the production MDI Agent platform on GCP (Cloud Run, BigQuery, Vertex AI); designed multi-agent orchestration workflows and data pipelines, democratizing self-service analytics and cutting time-to-insight by 30–50%.

Data Analytics & Dashboarding

Extracted, cleaned, and analyzed large-scale operational data including SLA lead times and IoT machine telemetry (2.6M+ logs); built interactive dashboards to track station bottlenecks and communicate actionable insights with cross-functional stakeholders.

Workflow Automation & Integration

Designed and deployed automated workflows consolidating data across Excel, Google Sheets, and cloud storage using Python, Apps Script, and Cloud Run; engineered automated shift-end reporting pipelines delivered directly via Google Chat and email.

Technical Consultation & Team Coordination

Provided technical consultation to cross-functional stakeholders on digital solutions; served as Team Coordinator aligning project deliverables, mentoring team members on data pipelines, and facilitating plant-wide Data Literacy training.

02 / 03

Olam Vietnam Food Processing

Digital Analyst · Jun 2025 – Sep 2025

02
3

Plants Deployed

10+ hrs

Saved / Week

−90%

Data Entry Errors

Connected Worker & Digitalization

Deployed the Poka Connected Worker platform across 3 manufacturing plants, digitizing standard operating procedures (SOPs) and boosting daily data collection and reporting efficiency by 40%.

Automated Reporting Pipelines

Built automated reporting workflows using Power Automate, Power BI, Forms, and VBA, delivering daily operational KPIs across Production, QA/QC, Maintenance, and HSE.

Digital Adoption & Consulting

Delivered digital adoption consulting and training across 15+ departments, standardizing operational workflows, saving 10+ hours/week, and reducing manual data entry errors by ~90%.

03 / 03

Bosch Vietnam (HCMC Plant)

Digital Transformation Intern · Jan 2025 – Jun 2025

03
2

Dashboards Built

1

Power App Maintained

100%

Technical Docs & Adoption

BI & KPI Dashboarding

Developed and maintained 2 Power BI dashboards (SQL, DAX) to track operational KPIs and support data-driven decision-making across plant operations.

Low-Code Tool Maintenance

Maintained and supported 1 internal Power Apps tool (including TGA Apprentice management) with integrated Power Automate workflows.

User Enablement & Documentation

Authored standard technical documentation and conducted user onboarding sessions to ensure sustainable long-term adoption of digital tools.

ENGINEERING PORTFOLIO

KEY PROJECTS

MDI Agent Multi-Agent Architecture on GCP
Flagship case study

MDI Agent — Manufacturing Data Intelligence Platform

Avery Dennison · Ongoing Production Development · 2025 – 2026

Plant operational data locked in disparate warehouse tables with slow manual shift reporting loops.

30–50%

Time-to-insight reduction

Sub-minute

Automated Chat Insights

GCP Cloud RunBigQueryVertex AIGoogle ADKFastAPIReact.
Read case study Live Trace Simulator Included
Supermarket Shelf Space Allocation Planogram and Benchmark

Supermarket Space Allocation & Shelf Optimization System

Published · CIRES 2025 · 2024 – 2025

Intuition-based supermarket shelf allocation causing profit leakage and inventory stagnation.

+19.9% profitability gain (2403 vs 2004), +119.8% turnover rate (0.68 vs 0.31), 2.01× heuristic compute agility. Published at CIRES 2025.

MILP PuLP/CBCBalanced ClusteringPython.
The Navigator Hybrid Career Recommender Architecture

The Navigator — Intelligent Career Path Recommender

Published · STDJE Journal · SR-ICYREB 2024 · 2023 – 2024

Students misaligned with industry skills, lack of personalized learning paths.

High-precision personalized curriculum recommendation (MAE: 0.1056, RMSE: 0.5866). Published in STDJE Journal & SR-ICYREB 2024 (2nd Prize UEL Young Scientist).

Collaborative Filtering Cosine SimilarityContent-Based KMeansKSA / Bloom's Taxonomy.
Mobile Banking Sentiment and Topic Analytics Engine

Mobile Banking Sentiment & Topic Analytics Engine (MyVIB)

Published · iCEBD 2025 · Graduation Thesis · 2024 – 2025

Unstructured customer app reviews unmonitored at scale.

Quantified user dissatisfaction (NPS: -19.9, CSAT: 42.5), isolating technical latency as primary churn driver. Published at iCEBD 2025 (Thesis Grade: 10/10).

SeleniumfastText 5-tier classificationLDA topic modelingCRISP-DMPython.
ACADEMIC & TECHNICAL CREDENTIALS

PUBLICATIONS & SKILLS

Peer-Reviewed Publications

CIRES 2025 · Lao Dong Publishing · 2025
Mathematical Optimization

Profit-Driven Space Allocation in Supermarkets: Comparing MILP and Heuristic Models

Formulated a Mixed-Integer Linear Programming model combined with Balanced Clustering for store planograms, evaluating profitability (2,403.05 vs 2,004.33) and inventory turnover (0.68 vs 0.31) on AEON Mall Vietnam dataset.

iCEBD 2025 Proceedings · 2025
Graduation Thesis 10/10 A+PDF

A New Approach of Sentiment Analysis and Topic Modeling for User-Generated Content in Mobile Banking Applications

Built an automated NLP ETL pipeline applying fastText 5-tier sentiment classification and LDA topic modeling on 6,700+ reviews, quantifying customer satisfaction indexes (NPS: −19.9, CSAT: 42.5).

SR-ICYREB 2024 · 2024
2nd Prize UEL Young ScientistPDF

Recommendation System for Personalized Learning Paths in Career Development

Engineered a hybrid recommendation engine combining Collaborative Filtering and Content-Based Filtering across 7,000+ job-skill pairs for personalized career curriculum guidance.

Science & Technology Development Journal (STDJE) · 2024
Journal PublicationDOI

Personalized Learning Paths Recommendation System with Collaborative Filtering and Content-Based Approaches

Formulated mathematical frameworks integrating KSA taxonomy and Bloom's Taxonomy for adaptive educational course module path generation (MAE: 0.1056, RMSE: 0.5866).

Formal Education

University of Economics and Law

VNU-HCM
Bachelor of Business: E-Commerce
Cumulative GPA8.54 / 10

Focus: Business Data Analytics, Recommendation Systems, Applied Mathematical Modeling, and Information Systems Architecture.

Core Tech Capabilities

Cloud & AI Engineering
GCP (Cloud Run, Vertex AI, BigQuery)Google ADKMulti-Agent SystemsDockerIAM / IAP
Programming & Backend
PythonFastAPISQL (BigQuery, PostgreSQL, SQL Server)DAXPlaywrightApps Script
Data & Mathematical Modeling
MILP (PuLP)fastTextLDA Topic ModelingKMeans ClusteringEDA & Metrics Modeling
BI & Enterprise Automation
Power BI (Cloud/DAX)Power AppsPower AutomateLooker StudioData Governance

Honors & Academic Awards

  • 2nd Prize: UEL Young Scientist Research Award (2023–2024)
  • Semi-Finalist: Eureka National Scientific Research Award (2024)
  • Academic Scholarship: Merit-based Scholarship (2021–2023)
  • Best Poster Design: MIS Faculty UEL (2024)

Professional Certifications

  • Google: Business Intelligence, Data Analytics & Project Management
  • Saint Louis University: Data Visualization & Advanced Reporting
  • English Proficiency:TOEIC L&R 885/990 · S&W 320/400