Machine Learning Engineer

Senior Machine Learning Engineer

Location: District 3, Ho Chi Minh City, Viet Nam

Work model: Onsite

Employment Type: Full-time

1. Role Overview

You will typically own one or two focus areas at a time, based on team priorities:

Search and Ranking

Design relevance features and ranking scores that directly improve search

result quality for users.

Apply learning-to-rank (LTR) techniques using LightGBM or XGBoost with

listwise/pairwise objectives, optimizing for NDCG and MRR.

Build and maintain embedding-based retrieval pipelines: generate query and

item embeddings, index them for approximate nearest neighbor (ANN) search,

and combine semantic retrieval with lexical matching for hybrid search.

Run offline evaluations and iterate on scoring logic to measurably improve

precision and recall.

Collaborate with product and engineering to validate ranking changes through

A/B tests and online metrics (CTR, conversion rate, engagement).

Recommendations and Personalization

Build recommendation scoring and ranking layers for use cases such as

similar items, similar users, and next-best suggestions.

Apply collaborative filtering, content-based filtering, or hybrid approaches

depending on data availability and use case.

Design recommendation strategies that adapt across platform surfaces

(search results, product detail pages, homepage feeds, email, push

notifications) using shared user/item representations with surface-specific

ranking.

Deliver personalization outputs that integrate into the product experience.

2. Key Responsibilities

Machine Learning Engineering

Build and improve ML models using scikit-learn, LightGBM, and XGBoost.

Apply transformer-based models (BERT, SentenceBERT) and sequence

models (RNN, LSTM, GRU) when appropriate.

Generate and fine-tune embeddings for semantic search and

recommendation retrieval. Work with vector similarity metrics (cosine similarity,

dot product) and ANN indexing (FAISS, pgvector, or similar).

Perform practical model improvement: feature selection, hyperparameter

tuning, thresholding, calibration, and error analysis.

Define and track offline evaluation metrics for each use case. Prevent quality

regressions with repeatable evaluation runs.

Be mindful of latency and serving constraints when designing models for real-

time search and recommendation touchpoints.

Feature Engineering

Create feature sets from behavioral and temporal signals: recency/frequency

patterns, rolling windows, and cohort-style aggregates.

Apply clean feature definitions and validation checks to prevent leakage and

instability.

Maintain clear feature documentation and versioning across iterations.

Code Quality and Testing

Own tickets end-to-end: clarify requirements, implement, test, and deliver

production-ready pull requests.

Write and maintain unit tests, integration tests, and API/contract tests using

pytest.

Create lightweight evaluation regression tests so model quality does not

degrade silently.

Containerized Development

Package ML services into Docker containers for consistent local and staging

validation.

Resolve Python dependency conflicts through version pinning, reproducible

builds, and systematic debugging.

Run services locally and in staging to verify correctness before handoff,

without relying on other teams for basic environment setup.

3. Required Qualifications

English communication requirement.

5+ years of professional experience building and shipping ML models in a product

environment.

Strong Python engineering skills for production ML code, including API

development with FastAPI or Flask, and GraphQL when needed.

Hands-on experience with scikit-learn, LightGBM, and XGBoost for structured data

problems.

Working knowledge of transformer models (BERT, SentenceBERT) and sequence

models (RNN, LSTM, GRU).

Solid data processing skills with NumPy and pandas.

Testing discipline with pytest, including fixtures and integration tests.

Strong Docker skills: building images, managing dependencies, and debugging

runtime issues in containers.

Ability to work independently: take a ticket, implement, test, and deliver without

constant coordination.

Hands-on experience with Elasticsearch or OpenSearch for building and tuning

search and relevance workflows.

4. Preferred Qualifications

Practical experience with tool orchestration and workflow automation.

Familiarity with MLOps practices: experiment tracking, model versioning, or CI/CD

for ML pipelines.

Experience with learning-to-rank (LTR) methods, ranking objectives, and evaluation

metrics such as NDCG, MRR, and MAP.

Hands-on experience with embeddings for retrieval: generating embeddings,

vector similarity search, and ANN indexing tools (FAISS, pgvector, Qdrant,

Pinecone, or similar).

Experience with recommendation techniques: collaborative filtering, content-based

filtering, two-tower architectures, or hybrid approaches.

Awareness of latency-performance tradeoffs when serving models in real-time search and recommendation systems.