Language systems2024—25
01
Data Warehouse+
Adaptive active learning for large-scale unstructured data, with BERT fine-tuning and production-ready data pipelines.
−38%annotation effort while maintaining model performance
AI researcher & engineer · HCM, Vietnam
I turn research ideas into AI systems that work in the real world.
My work connects deep learning research in video and signal processing with production-ready multimodal AI, language models, and real-time computer vision.
11Peer-reviewed
publications
publications
8dBPSNR gain in
reconstruction
reconstruction
38%Less annotation
effort
effort
PhDVideo coding &
deep learning
deep learning
01 / Selected work
From model research
to measurable impact.
Selected systems spanning language, vision, and real-time deployment.
Multimodal retrieval2024
02
ZumenAI
AI-powered CAD management with intelligent visual retrieval, automated tagging, and document classification.
+30%faster CAD image retrieval on average
Applied vision2023—24
03
AnzenAI
Autonomous safety analysis for construction imagery and video, built to generate contextual safety conditions on-site.
Real-time AI2023
04
Real-time safety monitoring
Low-latency accident-risk estimation with immediate on-site Patlite and email alerts.
View three more projects +
RoleplayEngine
Speech-to-text and BERT sentiment analysis for consistent sales-training evaluation.
Video coding research
Deep learning methods delivering up to 18% bitrate reduction against the VVC standard.
Compressive sensing
High-quality reconstruction for resource-limited surveillance devices, outperforming prior work by up to 6dB PSNR.
02 / Research
A body of work in
visual intelligence.
11 peer-reviewed works across video coding, compressive sensing, image enhancement, and computer vision. Filter the record by venue.
Computational Visual Media · Journal
JVCSR+: Adaptively Learned Video Compressive Sensing Reconstruction with Joint in-Loop Reference Enhancement and Out-Loop Super-Resolution
Data Compression Conference · DCC
Temporal Down-sampling based Video Coding with Frame-Recurrent Enhancement
IEEE MMSP
Object detection and tracking aided SLAM in image sequences for dynamic environment
MultiMedia Modeling · MMM
JVCSR: Video Compressive Sensing Reconstruction with Joint In-Loop Reference Enhancement and Out-Loop Super-Resolution
IEEE ISCAS
An 81.92Gpixels/s Fast Reconstruction of Images from Compressively Sensed Measurements
IEEE MMSP
SpeedDeblur: A Framework to speed up CNN-based Deblurring for HEVC compressed video
CVPR Workshops
Deep Learning based Spatial-Temporal In-loop filtering for Versatile Video Coding
IEEE Transactions on Industrial Informatics · Journal
CSIE-M: Compressive Sensing Image Enhancement Using Multiple Reconstructed Signals for Internet of Things Surveillance Systems
IEEE MMSP
Bi-directional intra prediction based measurement coding for compressive sensing images
IEEE Access · Journal
Deep Learning-Based Luma and Chroma Fractional Interpolation in Video Coding
International Symposium on AI & Robotics
iCNN: A Convolutional Neural Network for Fractional Interpolation in Video Coding
03 / Journey
Research depth,
industry range.
A career shaped in Japan and Vietnam, moving between academic research and end-to-end AI product development.
Eigyo-Mfg Co. Ltd. · Japan
AI Engineer
Owned end-to-end AI workflows from training and evaluation to deployment and monitoring; fine-tuned language models, built active learning systems, and mentored two interns.
Hazama Ando Corp. · Japan
AI Engineer, contractor
Engineered autonomous construction-safety systems, integrated vision AI on edge devices, and partnered with iOS and backend teams for large-scale deployment.
International venues
Reviewer
Reviewed video coding research for IEEE TCSVT and IEEE Access, and supported reviews for ICCV, CVPR, WACV, BMVC, VCIP, and OJSP.
Hosei University · Japan
Visiting Researcher
Developed machine-vision-oriented video coding approaches, reducing bitrate by up to 20% against state-of-the-art codecs.
JST PRESTO · Japan
Research Assistant
Researched deep learning-based compressive sensing for batteryless, wireless surveillance, delivering +8dB PSNR and 18% bitrate reduction.
04 / Capabilities
Across the full
AI system.
Research & modeling
Deep learning · Video coding · Image processing · Object detection · OCR · Multimodal AI · Active learning
Frameworks & languages
PyTorch · Hugging Face · scikit-learn · OpenCV · Python · C/C++ · MATLAB
Production & edge
ONNX Runtime · TensorRT · DeepStream · GStreamer · Docker · FastAPI
Data & cloud
PostgreSQL · pgvector · AWS EC2, S3, RDS · SageMaker · Git · Linux
05 / Recognition
Awards & support.
Doctoral Research Grant
Hosei University
Vietnam Friendship Scholarship
Hosei University & Taisei Corporation
JASSO Scholarship
Japan Student Services Organization
Top 5, graduating class
VNU-HCM UIT · Computer Science
Top Achiever Scholarship
Top 8% · VNU-HCM UIT
Honors Program Scholarship
VNU-HCM UIT
06 / Contact
Let’s make difficult ideas
useful.
Open to conversations about AI research, applied machine learning, computer vision, and production AI systems.
kimchi.phamdo@gmail.com ↗