NGUYEN LAN HUONG

# adtqnz@r.postjobfree.com +84-946****** § huonglarne

EXPERIENCE

ML engineer

* Moirai z 2022

• Improved and deployed an audio match-

ing software for movie-audio process-

ing companies.

AI engineer

* VTON z 2022

• Developed and deployed a virtual

clothes try-on web application.

AI engineer

* Aimesoft z 2021

• Improved the recognition algorithm

for an employee face recognition soft-

ware.

• Participated in R&D projects:

face animation according to voice

recording, image captioning in Viet-

namese.

AI research assistant

* USTH ICT Lab z 2020

• Participated in lung nodule segmenta-

tion research project.

LANGUAGES

• English: Fluent - 8.0 IELTS

• Vietnamese: Mother tongue

EDUCATION

B.S. in Information and

Communication technology

* University of Science and

Technology of Hanoi z 2018 - 2021

• Training language: English

• GPA: 3.7/4.0

• Degree classification: Very good

SKILLS

Machine learning/Deep learning

• Computer vision:

• Face recognition.

• Object detection and segmentation.

• Content transfer (motion, style etc).

• Audio processing:

• Voice recognition.

• Audio fingerprinting.

• Keyword spotting.

• Natural language processing:

• Vision-language models.

• Deep learning:

• Network implementation.

(Pytorch, Tensorflow, MXNet)

• Model training and parameter tuning.

(Optuna, Pytorch Lightning)

• Knowledge distillation / student-teacher training. MLOps

• Software engineering:

• CI/CD pipelines.

• Test-driven development.

• Deployment:

• Docker.

• FastAPI.

• Torchserve. Triton.

• SQL.

• AWS. GCP.

PROJECTS

Virtual clothes try-on web application

• Overview: Led a small team that laid the first bricks for a brand-new AI product.

Contributed to the roadmap and vision of the product.

• Technologies: Self-attention GAN. FastAPI. Docker.

• Contributions:

• Researched existing AI solutions to the problem.

= Chose and deployed a suitable core model for the product.

• Designed an architecture of separate services for each component model.

Implemented APIs for serving deep learning models.

= Allowed independent development and maintenance.

• Designed a suite of unit and integration tests for each service.

= Helped constructing the CI/CD pipeline.

• Created a standardized and shareable development environment.

= Increased development team’s productivity.

Audio matching software

• Overview: Deployed a 3-year stalling project into production. Made extensive improvement to the performance and maintainability of the product.

• Technologies: Audio fingerprinting. Github actions. AWS Lambda.

• Contributions:

• Optimized parameters of machine learning algorithm.

= Improved the model’s test accuracy by 36%.

• Designed a suite of unit and integration tests.

= Ensured continuous integration for CI/CD pipeline.

• Refactored the legacy code base.

= Better system maintainability.

= Reduced cost for developing new features.

Face recognition software

• Overview: Researched and implemented solutions to improve the perfor- mance of an existing AI product.

• Technologies: ResNet. ArcFace.

• Contributions:

• Fine-tuned the pre-trained face recognition model to optimize for Asian faces.

= Better sensitivity when identifying Vietnamese employees and customers.

• Implemented adaptive threshold for face embedding search.

= Improved the model’s accuracy by 4.8%.

• Used face clusters search.

= Reduced recognition speed for large face database. Recommendation engine for book website

• Technologies: SQL. Scikit-learn.

• Contributions:

• Designed a database to store the website’s data about products and customers.

= Optimize for data processing and query.

• Classified users into interest groups and designed a ranking score.

= Developed the book recommendation algorithm.

Lung nodule segmentation research project

• Technologies: R-CNN. U-Net.

• Contributions:

• Preprocessed the lung CT scan dataset.

= Clean dataset with easy-to-process format for future research.

• Proposed and implemented a 3D approach for nodule segmentation.

= Laid ground to future research paths in the project.

Contact this candidate