[01]
AI AND MACHINE LEARNING
Python · PyTorch · TensorFlow · Scikit-learn · XGBoost · Pandas · NumPy
STACK: PYTHON + PYTORCH + TENSORFLOW
THEME
PHAT_NGUYEN
APPLIED AI ENGINEER
█
▶ APPLIED AI ENGINEER — VIETNAM◆ COMPUTER VISION OCR LLM AUTOMATION MLOPS▶ FPT UNIVERSITY — GPA 3.75 / 4.0◆ SPRINGER LNEE CO-AUTHOR — 2026▶ 4 N8N WORKFLOWS — E-COMMERCE AUTOMATION◆ SELECTED: CODEPROVE LICENSE PLATE RECOGNITION SECURE LOGIN SYSTEM PROCON MANAGER FLUX AI DCASE 2025 ANOMALOUS SOUND DETECTION BASELINE▶ BUILDING PRACTICAL AI PRODUCTS PEOPLE CAN USE◆ GITHUB: PHATNT2K5▶ APPLIED AI ENGINEER — VIETNAM◆ COMPUTER VISION OCR LLM AUTOMATION MLOPS▶ FPT UNIVERSITY — GPA 3.75 / 4.0◆ SPRINGER LNEE CO-AUTHOR — 2026▶ 4 N8N WORKFLOWS — E-COMMERCE AUTOMATION◆ SELECTED: CODEPROVE LICENSE PLATE RECOGNITION SECURE LOGIN SYSTEM PROCON MANAGER FLUX AI DCASE 2025 ANOMALOUS SOUND DETECTION BASELINE▶ BUILDING PRACTICAL AI PRODUCTS PEOPLE CAN USE◆ GITHUB: PHATNT2K5
C:\PORTFOLIO> LOADING_NGUYEN_THANH_PHAT
PHATNGUYEN
APPLIED AI ENGINEER
◆ APPLIED AI ENGINEER BUILDING PRACTICAL COMPUTER VISION AND LLM-POWERED PRODUCTS.
LIVE_AI_DEMO / ONNX_WASM
DIGIT_RECOGNIZER.EXE
● MODEL_LOADING
INPUT_BUFFER560×560 → 28×28
Draw one digit (0–9) with a mouse or touch. Confidence updates while you draw.
DRAW 0–9LOADING LOCAL CNN
PREDICTED_DIGIT
?
CONFIDENCE00%
0
0%
1
0%
2
0%
3
0%
4
0%
5
0%
6
0%
7
0%
8
0%
9
0%
Loading the local ONNX digit model.
Current visual estimate: none.
- CURRENT GPA
- 3.75 / 4.0
- RESEARCH PUBLICATION
- 1 Springer paper
- AI AUTOMATION INTERNSHIP
- 4 n8n workflows
- SELECTED APPLIED BUILDS
- 6+ projects
ABOUT.TXT
PNSUBJECT_01.JPGProfile.jpg — PhotosPhat Nguyen01
$ cat bio.txt
Artificial Intelligence undergraduate at FPT University (GPA 3.75/4.0) focused on building practical AI products.
I turn computer vision, LLMs, APIs, and automation workflows into reliable end-to-end solutions—from experimentation to deployment.
Artificial Intelligence · FPT University · Vietnam
BASED IN VIETNAM — OPEN TO APPLIED AI OPPORTUNITIES
- 3.75 / 4.0
- CURRENT GPA
- 1 Springer paper
- RESEARCH PUBLICATION
- 4 n8n workflows
- AI AUTOMATION INTERNSHIP
- 6+ projects
- SELECTED APPLIED BUILDS
$ ls ./focus — 4 ACTIVE
APPLIED AI
ACTIVEEND-TO-END PRODUCT SYSTEMS
COMPUTER VISION
ACTIVEDETECTION + OCR
LLM AUTOMATION
ACTIVEAPIS + WORKFLOWS
MLOPS
ACTIVESERVING + EVALUATION
AUDIO RESEARCH
LEARNINGBASELINES + METRICS
EDUCATION FOCUS
Applied AIComputer VisionMachine LearningNatural Language ProcessingAI Product Engineering
SKILLS.EXE
▓
[02]COMPUTER VISION AND OCR
YOLOv8 · YOLO11 · OpenCV · MediaPipe · Tesseract OCR · Image preprocessing · Face recognition · Anti-spoofing
STACK: YOLOV8 + YOLO11 + OPENCV
◈
[03]NLP, LLM, AND AUTOMATION
Transformers · Attention models · OpenAI API · OpenRouter · RAG · Prompt workflows · n8n · SerpAPI · Webhooks
STACK: TRANSFORMERS + ATTENTION MODELS + OPENAI API
▶
[04]BACKEND AND DATA
FastAPI · Flask · SQLAlchemy · Alembic · PostgreSQL · MySQL · MongoDB · REST APIs · JWT · Docker
STACK: FASTAPI + FLASK + SQLALCHEMY
◆
[05]FRONTEND AND PRODUCT
Next.js · React · TypeScript · Vite · Tailwind CSS · Framer Motion · Three.js · UI/UX prototyping
STACK: NEXT.JS + REACT + TYPESCRIPT
◐
[06]AUDIO AND RESEARCH
Librosa · Log-Mel · MFCC · Autoencoders · Anomaly detection · Experiment tracking · AUC · pAUC · WER · CER
STACK: LIBROSA + LOG-MEL + MFCC
[07] WORKING MODE
COLLABORATION
Business analysisProblem framingSystem designTechnical documentationProject planningCross-functional collaboration
PROJECTS.BAT
6 PROJECTS.BAT
$ ls ./selected_projects
PROJECT_PREVIEW
CODEPROVE
PUBLIC REPOSITORY AVAILABLE
PRJ_001 · MVP / ACADEMIC DEVELOPMENT
PURPOSE: Evaluate the complete AI-assisted coding process, not only the final answer.
SOLUTION: A full-stack platform that captures understanding, hypotheses, AI prompting, verification, testing, debugging, explanation, and reflection.
ROLE: AI product and full-stack contribution
RESULT: An evolving MVP connecting assessment workflows, AI guidance, testing, and progress tracking.
FastAPIPostgreSQLOpenAI APINext.jsTypeScriptTailwind CSSDocker
PRJ_001 · MVP / ACADEMIC DEVELOPMENT
PURPOSE: Evaluate the complete AI-assisted coding process, not only the final answer.
SOLUTION: A full-stack platform that captures understanding, hypotheses, AI prompting, verification, testing, debugging, explanation, and reflection.
ROLE: AI product and full-stack contribution
RESULT: An evolving MVP connecting assessment workflows, AI guidance, testing, and progress tracking.
FastAPIPostgreSQLOpenAI APINext.jsTypeScriptTailwind CSSDocker
CodeProve
AI Product / Full-stack
MVP / academic development
PROJECT_PREVIEW
LICENSE PLATE RECOGNITION
PUBLIC REPOSITORY AVAILABLE
PRJ_002 · COMPLETED PROJECT · 01/2025 – 04/2025
PURPOSE: Recognize vehicle license plates accurately in an automated parking workflow.
SOLUTION: A YOLOv8 plate detector and YOLO11n 36-class character model integrated with preprocessing, confidence scoring, visualization, and a Flask interface.
ROLE: Computer Vision Engineer
RESULT: Best plate detector mAP50: 99.49% · Best plate detector mAP50–95: 75.64% · Character recognizer mAP50: 91.5% · Character recognizer mAP50–95: 69.6% · Character classes: 36
PythonYOLOv8YOLO11nOpenCVPyTorchFlask
PRJ_002 · COMPLETED PROJECT · 01/2025 – 04/2025
PURPOSE: Recognize vehicle license plates accurately in an automated parking workflow.
SOLUTION: A YOLOv8 plate detector and YOLO11n 36-class character model integrated with preprocessing, confidence scoring, visualization, and a Flask interface.
ROLE: Computer Vision Engineer
RESULT: Best plate detector mAP50: 99.49% · Best plate detector mAP50–95: 75.64% · Character recognizer mAP50: 91.5% · Character recognizer mAP50–95: 69.6% · Character classes: 36
PythonYOLOv8YOLO11nOpenCVPyTorchFlask
License Plate Recognition
Computer Vision / OCR
Completed project
PROJECT_PREVIEW
SECURE LOGIN SYSTEM
CV PROJECT · LINK NOT PROVIDED
PRJ_003 · COMPLETED PROJECT · 06/2025 – 11/2025
PURPOSE: Build a secure login flow that verifies identity and rejects presentation attacks.
SOLUTION: A computer-vision pipeline combining face detection, liveness checks, YOLOv8-cls and LivenessNet anti-spoofing, and face recognition.
ROLE: Computer Vision Engineer
RESULT: CV-reported accuracy: 98% · Precision / recall / F1: 0.98 · Evaluation samples: 3,147 · APCER / NPCER: 2.45% / 2.44% · ACER / EER: 2.45% / 2.45% · EER threshold: ≈0.5061
PythonYOLOv8-clsLivenessNetOpenCVFace Recognition
PRJ_003 · COMPLETED PROJECT · 06/2025 – 11/2025
PURPOSE: Build a secure login flow that verifies identity and rejects presentation attacks.
SOLUTION: A computer-vision pipeline combining face detection, liveness checks, YOLOv8-cls and LivenessNet anti-spoofing, and face recognition.
ROLE: Computer Vision Engineer
RESULT: CV-reported accuracy: 98% · Precision / recall / F1: 0.98 · Evaluation samples: 3,147 · APCER / NPCER: 2.45% / 2.44% · ACER / EER: 2.45% / 2.45% · EER threshold: ≈0.5061
PythonYOLOv8-clsLivenessNetOpenCVFace Recognition
Secure Login System
Face Authentication / Anti-spoofing
Completed project
PROJECT_PREVIEW
PROCON MANAGER
CV PROJECT · LINK NOT PROVIDED
PRJ_004 · IN PROGRESS · 04/2026 – PRESENT
PURPOSE: Make complex construction contracts easier to review, track, and act on.
SOLUTION: A platform for document upload and parsing, AI extraction, milestone and approval workflows, and dashboards for progress, overdue tasks, and risks.
ROLE: AI Product Engineer / Frontend Developer
RESULT: An in-progress platform connecting contract intelligence with operational tracking and review workflows.
Next.jsPrismaPostgreSQLRedisDockerAWS S3
PRJ_004 · IN PROGRESS · 04/2026 – PRESENT
PURPOSE: Make complex construction contracts easier to review, track, and act on.
SOLUTION: A platform for document upload and parsing, AI extraction, milestone and approval workflows, and dashboards for progress, overdue tasks, and risks.
ROLE: AI Product Engineer / Frontend Developer
RESULT: An in-progress platform connecting contract intelligence with operational tracking and review workflows.
Next.jsPrismaPostgreSQLRedisDockerAWS S3
ProCon Manager
AI Product / Contract Intelligence
In progress
PROJECT_PREVIEW
FLUX AI
CV PROJECT · LINK NOT PROVIDED
PRJ_005 · IN PROGRESS · 04/2026 – PRESENT
PURPOSE: Deliver timely, understandable financial guidance inside a mobile banking experience.
SOLUTION: A mobile-first AI coach that reacts to transactions, classifies behavior, detects moments of need, and delivers nudges, product guidance, and RAG-grounded FAQ responses through an AI avatar.
ROLE: AI Product Engineer / Frontend Developer
RESULT: An in-progress product concept combining proactive coaching, safer product answers, and personalized recommendations.
RAGLLM WorkflowsBehavior ClassificationAI AvatarMobile-first UX
PRJ_005 · IN PROGRESS · 04/2026 – PRESENT
PURPOSE: Deliver timely, understandable financial guidance inside a mobile banking experience.
SOLUTION: A mobile-first AI coach that reacts to transactions, classifies behavior, detects moments of need, and delivers nudges, product guidance, and RAG-grounded FAQ responses through an AI avatar.
ROLE: AI Product Engineer / Frontend Developer
RESULT: An in-progress product concept combining proactive coaching, safer product answers, and personalized recommendations.
RAGLLM WorkflowsBehavior ClassificationAI AvatarMobile-first UX
Flux AI
AI Product / Personal Finance
In progress
PROJECT_PREVIEW
DCASE 2025 ANOMALOUS SOUND DETECTION BASELINE
PUBLIC REPOSITORY AVAILABLE
PRJ_006 · BASELINE EXPERIMENT
PURPOSE: Explore anomaly detection across machine and domain conditions using normal audio.
SOLUTION: A reproducible pipeline with log-Mel and MFCC features, convolutional autoencoders, latent-distance experiments, metrics, plots, predictions, reports, and slides.
ROLE: Experiment design, implementation, evaluation, and documentation
RESULT: The simple reconstruction baseline is near random on several machines, clearly documenting limitations and next steps.
PythonTensorFlowLibrosaScikit-learnNumPyPandasMatplotlib
PRJ_006 · BASELINE EXPERIMENT
PURPOSE: Explore anomaly detection across machine and domain conditions using normal audio.
SOLUTION: A reproducible pipeline with log-Mel and MFCC features, convolutional autoencoders, latent-distance experiments, metrics, plots, predictions, reports, and slides.
ROLE: Experiment design, implementation, evaluation, and documentation
RESULT: The simple reconstruction baseline is near random on several machines, clearly documenting limitations and next steps.
PythonTensorFlowLibrosaScikit-learnNumPyPandasMatplotlib
DCASE 2025 Anomalous Sound Detection Baseline
Audio ML / Research Experiment
Baseline experiment
VIEW GITHUB PROFILE
github.com/PhatNT2k5
▶
EXPERIENCE.LOG
APPLIED AI EXPERIENCE + EDUCATION
EXP_001
AI AUTOMATION INTERN / AI ENGINEER INTERN
EcomEasy
January 2026 – April 2026
Built AI automation workflows for e-commerce content operations and market research while collaborating across AI, backend, frontend, UI/UX, and delivery functions.
- ▶Developed four n8n pipelines for TikTok Shop, Shopee, Lazada, and market research
- ▶Integrated LLMs, SerpAPI, webhooks, APIs, and JavaScript nodes
- ▶Automated product copywriting, competitor research, trend analysis, and structured reports
- ▶Contributed to interface refinement and cross-functional delivery
- ▶Participated in LotusHack and Qwen Build AI Day
n8nLLM APIsSerpAPIWebhooksREST APIsJavaScript
EDU_001
FPT UNIVERSITY
Artificial Intelligence · Vietnam
CURRENT GPA
3.75 / 4.0
FOCUS AREAS
Applied AIComputer VisionMachine LearningNatural Language ProcessingAI Product Engineering
Study dates and campus are intentionally omitted until verified.
RESEARCH.DB
RESEARCH AND RECOGNITION
RES_001 · CO-AUTHOR · 1 MAY 2026
Optimizing YOLOv11n for Real-Time Object Detection: Leveraging Quantization and Model Optimization
The work studies YOLOv11n optimization for CPU-constrained deployment using weight quantization while retaining sensitive activations in FP32 to reduce resource consumption without significantly degrading accuracy.
- PUBLISHER
- Springer Singapore
- SERIES
- Lecture Notes in Electrical Engineering, Volume 1531
- PAGES
- 159–170
- DOI
- 10.1007/978-981-95-6111-7_13
$ cat achievements.log
Springer LNEE Co-author
2026 · VERIFIED
Top 100 Excellent Students
Fall 2025 · PROOF PENDING
Honorable Student of Trimester
SU24–SU25 · PROOF PENDING
CREDENTIALS.DIR
SELECTED LEARNING THAT SUPPORTS HOW I BUILD, DEPLOY, AND EVALUATE AI SYSTEMS
[01]SPECIALIZATION
MLOps & Model Serving
Coursera · July 2026
MLOpsModel ServingML MonitoringProduction ML
VERIFY CREDENTIAL ▶[02]SPECIALIZATION
AI Engineer Professional
Packt · June 2026
AI EngineeringDeep LearningAI AgentsModel Optimization
VERIFY CREDENTIAL ▶[03]SPECIALIZATION
Data Science Fundamentals
IBM · March 2025
PythonSQLData ScienceStatistics
VERIFY CREDENTIAL ▶[04]SPECIALIZATION
Software Development Lifecycle
University of Minnesota · May 2025
Software EngineeringAgileQualitySDLC
VERIFY CREDENTIAL ▶[05]COURSE
NLP with Attention Models
DeepLearning.AI · March 2026
NLPAttentionTransformersSequence Modeling
VERIFY CREDENTIAL ▶[06]COURSE
Containers & Kubernetes
IBM · October 2025
DockerKubernetesOpenShiftDeployment
VERIFY CREDENTIAL ▶CONTACT.SH
START A CONVERSATION ABOUT APPLIED AI
CONTACT INFO
- GITHUB
- github.com/PhatNT2k5
- Phát Nguyễn
- LOCATION
- Vietnam
SYSTEM STATUS
- PRIMARY ROLE
- Applied AI Engineer
- CURRENT
- Artificial Intelligence Undergraduate
- FOCUS
- CV · OCR · LLM AUTOMATION
- CONTACT
- EMAIL PREFERRED
$ ./open_mail_client.sh
LET'S TALK
Email is the direct contact channel for AI engineering opportunities, project collaboration, and research conversations. This site does not simulate form submissions.
[ EMAIL PHAT ]