CV
My current resume as an AI engineer/AI researcher.
Contact Information
| Name | Pham Thanh Trung |
| Professional Title | AI Researcher / Engineer |
| phmtrung2@gmail.com | |
| Phone | +84-845604327 |
| Location | Yongin, Gyeonggi-do |
Professional Summary
A product-oriented engineer with a strong ability to research AI (ML/DL), robust model training skills, and a visionary architecture design mindset. Excelled at designing, orchestrating, and implementing best-fit AI approaches to deliver production-ready solutions. Practical and highly experienced at conducting technical research to overcome development bottlenecks without trading off the delivery schedule.
Experience
-
2025 - Present Yongin City, Korea
AI Researcher - PhD student
Kyunghee University
- Research on quantum computing and multi-modal LLMs for advanced AI applications.
- Reduced VGG19 classification parameters from 3 million to 150 quantum parameters with a 0.025 improvement in image classification accuracy.
-
2024 - 2025 Ho Chi Minh City, Vietnam
Artificial Intelligence Engineer (Mid-level)
IMT Solutions
- Built LLM, MLOps, computer vision, and agentic AI systems for enterprise clients.
- Delivered an agentic KIE system with 94% peak accuracy and over 90% operational accuracy while reducing inference cost by more than 40%.
- Automated scanned PDF to DITA conversion, reducing digitization from weeks to 6-16 minutes per document.
- Improved machine translation BLEU score from 0.74/0.68 to 0.93 using multi-agent NLLB + Wikidata pipelines.
- Designed scalable AI architectures and coordinated implementation tasks.
-
2022 - 2023 Ho Chi Minh City, Vietnam
Artificial Intelligence Research Intern
GMO-Z.com RUNSYSTEM Joint Stock Company
- Proposed Fourier transform approaches for deepfake and anti-spoofing classification, improving F1 score from 0.86 to 0.91.
Education
-
2025 - Present Yongin, Republic of Korea
PhD combined program
Kyunghee University Global Campus
Artificial Intelligence
-
2020 - 2023 Ho Chi Minh City, Vietnam
Bachelor
FPT University Ho Chi Minh City
Artificial Intelligence
- Excellent graduate thesis (graded 9.2/10)
Projects
-
Key Information Extraction from Documents
2024-05 – 2025-02Team leader and main contributor.
- Applied multi-modal LLMs to extract invoice numbers, prices, and item names.
- Built an Azure Document Intelligence extraction system with CI/CD and layout recognition and labeling workflows.
- Reverse-engineered Azure custom model services and added few-shot layout classification and multi-agent processing pipelines.
- Stack: Azure Document Intelligence, LangChain, FastAPI, PyTorch, Hugging Face, Transformers/TGI, Qdrant, Gemini API, OpenAI API, Azure Pipelines, Docker.
-
Pali ancient language of the Buddhism Tipitaka machine translator
2023-02 – 2025-09Personal project.
- Mined and curated 120K Pali-English parallel sentence pairs using LaBSE and LASER.
- Fine-tuned NLLB-200 (1.3B and 3B) and BGE-M3 for Pali-Vietnamese-English tasks.
- Improved BLEU score from 0.5542 to 0.7837 and deployed optimized models.
- Stack: PyTorch, FastAPI, Hugging Face TGI/Transformers, wandb, Docker, LASER, BGE-M3, LangChain, FAISS.
-
Research on LLMs to enhance domain-driven neural machine translation
2024-01 – 2025-02Team leader and main contributor.
- Integrated RAG, chain-of-thought, and multi-agent systems to boost translation without retraining at deployment.
- Fine-tuned NLLB on domain-specific datasets and expanded the dataset by 60% using automated data mining.
- Stack: FastAPI, Gradio, NLLB, Hugging Face Transformers, LangChain.
-
Medical vision Q&A LLM system
2025-02 – 2025-03Personal project.
- Fine-tuned InternVL2.5 on VQA-RAD and built a ChatGPT-style UI with auth and attachments.
- Built and deployed LLM serving APIs, microservices, and database-backed backend.
- Stack: FastAPI, Hugging Face TGI/Transformers, SQLAlchemy, PyTorch, Streamlit.
-
Quantum machine learning research
2025-02 – PresentMain researcher.
- Investigated applying quantum computing capabilities to AI and ML.
- Implemented and optimized hybrid quantum-classical deep learning architectures.
- Stack: Pytorch, PennyLane.
-
Fortune telling chatbot
2025-09 – 2025-10Personal project.
- Designed a multi-modal RAG knowledge ingestion pipeline to fetch contextual knowledge tailored to user story and persona.
- Built a generalized AI framework for tarot, numerology, astrology, and similar domains without domain-specific fine-tuning.
- Implemented LLM serving APIs, knowledge databases, ingestion services, user databases, and authentication service.
- Built a basic chat interface for end users and CI/CD deployment.
- Stack: FastAPI, Hugging Face TGI, OpenAI APIs, Gemini APIs, Qdrant, SQLAlchemy, Pytorch, Gradio.
-
Automated scanned PDF to DITA converter with generative AI
2024-01 – 2025-02Team leader, main contributor.
- Researched multi-modal language models to convert scanned PDF manuals into structured DITA format.
- Implemented a multi-agent system to improve extraction accuracy.
- Stack: FastAPI, Pytorch, Gemini API, OpenAI API, LangChain.
-
Employee training chatbot system
2024-02 – 2025-05Main contributor.
- Implemented a document ingestion system using multi-modal LLMs to structure internal training documents in a vector database.
- Built an RAG chatbot with self-hosted LLMs for internal policy Q and A and automatic ingestion of policy updates.
- Stack: FastAPI, Qdrant, Streamlit, Docker.
-
Vitexco Exemplar-based Video Colorization using Vision Transformer
2023-01 – 2023-08- Applied vision transformers for video colorization with reference images while maintaining temporal color consistency across frames.
- Created large-scale video datasets for training.
- Conducted training experiments to optimize model architecture and performance.
- Set up training servers and monitored model parameters.
-
eKYC modules development
2022-11 – 2023-02Researcher, developer.
- Created and curated large-scale datasets for training eKYC modules.
- Designed and applied deep learning architectures to solve eKYC challenges.
- Stack: Pytorch, TensorFlow, OpenCV, MediaPipe, PyVista.
-
Cosmetic products recommender system based on facial skin images
2023-02 – 2023-02Team leader, main contributor.
- Built DL submodules for acne detection and classification and skin moisture regression.
- Created, preprocessed, and augmented facial skin image datasets.
- Quantized models, re-implemented interfaces, and converted models to ONNX for deployment.
- Developed a Flutter mobile app and deployed AI APIs and databases.
- Stack: FastAPI, Flutter, Google Cloud Platform, YOLOv8.
-
3D reconstruction with deep learning
2022-07 – 2022-07Team member.
- Built deep-learning 3D reconstruction for environment virtualization and single-camera motion capture.
- Implemented facial and body motion capture, blendshape inference, and motion retargeting modules.
- Created large-scale 3D datasets for blendshape inference training.
- Stack: Pytorch, TensorFlow, Unreal Engine 4 MetaHuman, Blender.
-
Solving eight-queens problem with genetic algorithm
2017-08 – 2017-08Personal project.
- Implemented a custom genetic algorithm framework in Pascal.
Skills
Publications
-
2025 SwinTExCo: Exemplar-based video colorization using Swin Transformer
Expert Systems with Applications
Achievements
- 100% Scholarship at Kyunghee University for PhD combined program.
- Champion prize, FPT Edu Research Festival finale 2023.
- 100% Scholarship at FPT University HCM.
- Third prize, FPT Education Mathematics Olympic 2021.
- Silver medal, Traditional Olympic 30/4 Contest 2018 (informatics).