hội tụ tại AIPractix is pleased to announce the successful mentorship of Ms. Nguyen Duong Kim Anh, a Computer Science student, who has completed her graduation thesis with an Excellent evaluation.
The thesis, titled:
DESIGN AND IMPLEMENTATION OF A DEEP LEARNING–BASED TEXT–IMAGE FEATURE REPRESENTATION SYSTEM FOR FASHION PRODUCT SEARCH FROM TEXT QUERIES
was highly regarded for its technical depth, modern approach, and strong real-world applicability, particularly through its direct application to FeatSearch, AIPractix’s fashion-focused AI search platform.
The Challenge: Searching Fashion Products from Text Is Hard
Fashion product search presents unique challenges that traditional keyword-based systems fail to address:
- User queries are often vague and subjective, describing style, mood, or appearance rather than exact product names
- There is a significant semantic gap between textual descriptions and visual product features
- Fashion search requires understanding both meaning and visual similarity
Ms. Nguyen Duong Kim Anh’s thesis addressed these challenges by learning a joint representation between text and images using deep learning techniques, enabling effective cross-modal retrieval for fashion products.
Core Contributions of the Thesis
The research focused on several key technical components:
- Designing a text–image feature representation system for fashion data
- Training deep learning models to map:
- Text queries
- Product images
into a shared embedding space
- Implementing a text-to-image retrieval pipeline, where users search using natural language and receive visually and semantically relevant fashion products
The system was evaluated on real fashion datasets, demonstrating strong retrieval accuracy and practical usability.
AIPractix’s Mentorship Role
Throughout the thesis, AIPractix served as a technical mentor, guiding the project with a strong emphasis on production-oriented AI design.
1. System Architecture and Model Design
AIPractix supported the student in:
- Structuring the overall system architecture
- Separating text encoding, image encoding, and joint embedding learning
- Ensuring the design aligned with real-world fashion search requirements
2. Bridging Research and Product Development
A key strength of the project was its direct integration into FeatSearch, AIPractix’s AI-powered fashion search engine.
This allowed the research to be:
- Validated on real product data
- Evaluated under practical constraints such as scalability and performance
- Positioned as a production-ready component, not just an academic prototype
3. Practical and Responsible AI Thinking
Beyond model performance, the mentorship emphasized:
- Real user search behavior
- System robustness and maintainability
- The importance of measurable impact in applied AI systems
Results and Academic Recognition
- The thesis received an Excellent grade
- The evaluation committee highlighted:
- The suitability of deep learning for text–image fashion search
- Clear system design and evaluation methodology
- Strong relevance to real-world applications
- The research contributed directly to the text–image retrieval module of FeatSearch
Commitment to Education and Practical AI
AIPractix remains committed to:
- Mentoring Computer Science students
- Supporting:
- Graduation theses
- Applied AI research
- Industry–academia collaboration
Our core mentoring and research focus areas include:
- Computer Vision and Multimodal AI
- Text–Image Retrieval
- Semantic Search for E-commerce
- AI Platforms and APIs
We strongly believe that:
AI research creates real value only when it is validated through practical systems and real users.
Congratulations
AIPractix congratulates Ms. Nguyen Duong Kim Anh on her outstanding academic achievement and wishes her continued success in her future career in artificial intelligence and software engineering.
We look forward to mentoring and collaborating with more students and institutions to bring practical, responsible AI into real-world applications.
AIPractix – Where Practical AI Powers Real Life
Vietnamese 
English