End-to-end AI pipelines
Data preprocessing, training, evaluation, explainability, and demo deployment.
Open to AI / Computer Vision internships & research roles
Computer Vision &
Applied AI Student
I am Nguyen Trong Bach, an ICT undergraduate at the University of Science and Technology of Hanoi. I build complete AI pipelines — from data preprocessing and model training to evaluation, explainability, and demo deployment — with a focus on Computer Vision, Medical Image Analysis, OCR, and Applied NLP.
USTH · B.Sc. ICT
AI / Computer Vision
Open to Internship
Master 2IS Direction
01 — About
About Me
I am an ICT undergraduate at the University of Science and Technology of Hanoi (USTH) with a strong focus on Computer Vision and Applied AI. My work centers on building complete AI pipelines, from data preprocessing and model training to evaluation, explainability, and demo deployment.
I am particularly interested in medical image analysis, object detection, video understanding, document AI, and robust real-world AI systems. I enjoy taking a project from an open dataset all the way to a working interface that a non-technical user can try — including model export, lightweight backends, and mobile demonstration.
I am preparing to continue my academic path through the Master 2IS program at Toulouse Capitole 1 University, where I aim to deepen my expertise in information systems and applied AI for real-world digital solutions.
Practical, deployment-aware
FastAPI / Streamlit / Gradio / ONNX export — from notebook to real interface.
Research-oriented mindset
Comparative experiments, ablations, metrics, confusion matrices, Grad-CAM.
02 — Education
Education
B.Sc. in Information and Communication Technology
Dec 2023 – Aug 2026 (expected)University of Science and Technology of Hanoi (USTH)
13.19 / 20
GPA
6.5
IELTS
Aug 2026
Expected Graduation
Coursework spans the full software and AI stack: algorithms, systems, statistics, and modern deep learning. My academic project work is concentrated in Computer Vision, NLP, and applied AI pipelines.
03 — Next Step
Future Master’s Direction
Target Program · 2026 Intake
Master 2IS — Toulouse Capitole 1 University
I am preparing to continue my graduate studies in the Master 2IS (Information Systems & Software Engineering) program at Toulouse Capitole 1 University. The program aligns closely with my interests in information systems, applied AI, software systems, data-driven applications, and intelligent digital solutions.
My goal is to strengthen my foundations in software engineering and information systems while continuing to specialize in computer vision and applied AI, contributing to research and industrial projects that bring AI from prototypes into reliable, deployable systems.
04 — Research
Research Interests
Computer Vision
Image classification, detection, segmentation, and visual recognition.
Medical Image Analysis
Dermoscopy, diagnostic classification, and explainable medical AI.
Object Detection
YOLO-family models, real-time inference, and bounding-box pipelines.
Video Understanding
Temporal models, action recognition, and pose-based pipelines.
Deep Learning
CNNs, Transformers, fine-tuning, and architecture comparison.
Efficient Real-time Inference
Model export, ONNX conversion, and lightweight deployment.
Data-centric AI
Data curation, class imbalance, augmentation, and quality auditing.
Model Robustness & Interpretability
Grad-CAM, error analysis, and reliability under distribution shift.
OCR & Document Understanding
Receipt parsing, layout-aware models, and end-to-end document AI.
Applied NLP
Transformer-based NER, information extraction, and inference workflows.
05 — Skills
Technical Skills
Programming
Machine Learning & Deep Learning
Computer Vision
NLP & Document AI
Deployment
Development Workflow
06 — Work
Featured Projects
Seven selected projects across Computer Vision, Medical AI, Document AI, NLP, Security, and Mobile — each shipped with code, evaluation, and (where relevant) a working demo.
B.Sc. Thesis · Medical AI
Skin Lesion Classification
A thesis-level deep learning project for 7-class dermoscopic skin lesion classification using public datasets (HAM10000, ISIC 2018). Compares CNN and Transformer backbones, addresses class imbalance, and includes evaluation, confusion-matrix analysis, prediction export, and Grad-CAM explainability.
88.4%
Best single-model accuracy
0.831
Macro F1
0.980
Ensemble AUC
- Backbones: ResNet50, DenseNet121, EfficientNet-B0, Swin-Tiny
- Loss: cross-entropy, weighted CE, focal loss
- Grad-CAM explainability + Streamlit / Gradio demo
- Designed for local RTX training with Colab fallback
Video & Pose · Computer Vision
Figure Skating Action Recognition
A data-first action-recognition project for figure skating elements, exploring both pose-based and video-based approaches. Includes 3D skeleton sequences, temporal modeling, video clip preprocessing, manifest building, and baseline evaluation.
- Data sources: FS-Jump3D, FSD-10, FineFS (planned / referenced)
- Pose-based and lightweight video-based baselines
- End-to-end preprocessing, tensor conversion, training, testing scripts
- Pose baseline currently outperforms the lightweight video baseline
Object Detection · Mobile
Smoking Detection System
A practical computer-vision system for detecting smoking-related visual cues in images, video, and live camera streams. Covers YOLO training, video inference, ONNX export, and an Android application for user-facing demonstration.
- Fine-tuned YOLO11 for cigarette, cigar, and smoke classes
- Single image, video file, and webcam / live camera inference
- Model export to ONNX for future mobile deployment
- Android prototype built in Java with Android Studio
Document AI · OCR
SROIE Receipt OCR & Information Extraction
A document-AI project for OCR and key information extraction from receipt images. Compares OCR-driven rule-based extraction, LayoutLMv3 token classification, and DONUT-style end-to-end document parsing for structured field extraction.
- Full receipt understanding pipeline
- OCR preprocessing and text-box extraction
- Layout-aware information extraction
- End-to-end document parsing exploration
- Field-level evaluation: accuracy, F1, normalized edit distance
NLP · Named Entity Recognition
IT Job Description Entity Extraction
A Transformer-based named-entity-recognition project for extracting structured information from unstructured IT job descriptions — role, skill, location, experience, salary — with an inference workflow exposed through a FastAPI backend.
- Compares BERT, DistilBERT, and RoBERTa backbones
- Extracts career-related structured entities
- FastAPI-based inference workflow
Security · Cryptography
Secure Online Banking Session
A security-focused prototype demonstrating secure online banking sessions: TLS 1.3 communication, password hashing, session protection, CSRF protection, and encrypted traffic verification with Wireshark.
- TLS 1.3 communication
- bcrypt password hashing
- HttpOnly cookie-based sessions
- CSRF protection
- Wireshark verification of encrypted traffic
Mobile · Android
Mobile Application Final Project
An Android application focused on practical mobile UI implementation, activity management, data handling, and app structure in Java using Android Studio.
- UI design and activity lifecycle handling
- Data handling and structured app architecture
- API integration
No projects in this category yet.
07 — Coursework
Selected Coursework
- Machine Learning
- Deep Learning
- Digital Image Processing
- Natural Language Processing
- Probability and Statistics
- Data Structures and Algorithms
- Object-Oriented Programming
- Distributed Systems
- Computer Networks
- Cryptography
- Database Systems
- Mobile Programming
08 — Experience
Experience
IT Intern — FUJILOGIC Technology Co., Ltd.
Oct 2025 – PresentHanoi, Vietnam
Delivered front-line technical support for software configuration, troubleshooting, routine maintenance, internal engineering workflows, environment setup, and technical documentation.
IT Intern — C+ Technology JSC
Feb 2025 – Aug 2025Hanoi, Vietnam
Supported IT operations, software deployment, environment configuration, integration testing, issue reproduction and resolution, documentation, and team communication in a professional software environment.
09 — Activities
Activities
USTH Coders Club — Marketing Team (Content & Design)
Oct 2023 – Jun 2025Produced promotional content and visual materials for programming-related activities, completed internal coding workshop activities, and collaborated closely with student organizers to support club operations.
10 — Curriculum Vitae
View My CV
PDF preview — for the best experience on mobile, tap Open in New Tab.
11 — Contact
Let’s Get in Touch
Interested in collaboration, an internship, or research opportunities?
I’m actively looking for AI / Computer Vision internships and research collaborations. The fastest way to reach me is by email — I usually reply within a day.
Send Me an Email