Tran Minh Khoa, data scientist

Tran Minh Khoa

Senior Data Scientist

Ho Chi Minh City, Vietnam · UTC+7 · 8 years of experience

Tran focuses on turning messy data into tested models and useful decisions.

Request a shortlist

Why this profile fits

Decision-first problem framing

Starts with the decision, target metric, and data limits before choosing a model or analytical method.

Validated modeling

Uses holdouts, experiment design, and error analysis to separate a useful signal from a convincing notebook.

Clear operational handoff

Documents assumptions, monitoring needs, and the path from analysis to a decision or production workflow.

Relevant toolkit

PythonSQLscikit-learnPyTorchdbt

Delivery evidence

Selected work

Embedding-Based Recommender Migration

Consumer e-commerce marketplace

Replaced a legacy collaborative-filtering engine with a two-tower retrieval model and pgvector approximate nearest-neighbor search feeding a gradient-boosted ranker. Add-to-cart conversion rose 23% and long-tail catalog coverage nearly doubled, validated through a staged A/B test across 12 million monthly users.

PythonPyTorch

Real-Time Personalization Feature Store

Streaming and media platform

Designed a low-latency feature store and streaming pipeline so ranking models could react to in-session behavior within seconds rather than hours. Cut feature staleness from overnight to under 30 seconds and improved session watch-time by 11% on personalized rails.

PythonPolars
View more profile detail →

Relevant experience

Career timeline

Senior Data Scientist, Personalization and Recommendations

Feb 2022–Present

Regional E-Commerce Marketplace (30M+ MAU) · Ho Chi Minh City, Vietnam (Remote)

  • Led the migration from item-to-item collaborative filtering to a two-tower retrieval and learning-to-rank stack, lifting homepage add-to-cart conversion 23% across 12 million monthly users.

Data Scientist, Recommendations and Growth

Apr 2020–Jan 2022

Consumer Fintech and Lifestyle Super-App · Ho Chi Minh City, Vietnam

  • Built the personalized offers and content-ranking models that grew in-app feed click-through rate 17% and drove a 9% lift in cross-sell of financial products.

Data Scientist

Jul 2018–Mar 2020

Digital Media and Ad-Tech Agency · Ho Chi Minh City, Vietnam

  • Built content-recommendation and audience-segmentation models that increased average pages-per-session 14% across a network of publisher sites.

Skill depth

Show experience in context.

VerbalCommunicationDomainUnderstandingProblemSolvingCodeQualitySystemDesignDeliverySpeed
Competency shapeAssessment across the same six dimensions used for every profile.

Relevant experience by skill

Python8 years
SQL8 years
Machine Learning8 years
Recommendation Systems6 years
A/B Testing & Experimentation6 years
PyTorch5 years
Snowflake5 years
MLflow4 years
See the complete skill index

Languages

PythonSQLRScalaBash

Frameworks

PyTorchTensorFlowscikit-learnLightGBMXGBoostStreamlitFastAPI

Libraries/APIs

pandasPolarsNumPySciPystatsmodelsLangChainHugging Face TransformersimplicitLightFMSHAP

Tools

MLflowdbtAirflowDockerGitWeights & BiasesFeastDVC

Paradigms

Recommendation SystemsLearning-to-RankCollaborative FilteringA/B Testing & ExperimentationCausal InferenceForecastingMLOpsReal-Time Personalization

Platforms

SnowflakeBigQueryDatabricksAWS SageMakerGCP Vertex AIKubernetes

Storage

PostgreSQLpgvectorRedisBigQuerySnowflakeAmazon S3

Other

Vector SearchSemantic SearchRAGCold-Start ModelingFeature StoreGuardrail Metrics

Education

Formal background

Bachelor of Engineering, Computer Science

Ho Chi Minh City University of Technology (VNU-HCM) · 2014–2018 · Distinction, GPA 3.6/4.0

Master of Science, Data Science

John von Neumann Institute, VNU-HCM · 2019–2021

Credentials

Certifications

Databricks Certified Machine Learning Professional

Databricks · 2025 · Certified

SnowPro Advanced: Data Scientist

Snowflake · 2024 · Certified

Communication

Languages

Vietnamese

Native

English

Professional

Japanese

Conversational

More examples

Similar Data Scientist profiles.