resume
I hope this page give you more knowledge about my professional life.
Basics
| Name | PHAN Vỹ Hào |
| Label | Data Scientist |
| vyhao03@gmail.com | |
| Phone | (84) 947-368-311 |
| Url | https://vyhaophan.github.io |
| Summary | I have more than 2 years of experience in data science, machine learning, and AI. My major is LLM models development and application in recruitment and office automation domain. I have also developed some projects in Computer Vision for tracking and counting people. |
Work
-
2024.08 - Now Data Scientist
Mylong Inc.
Developing an intelligent HR assistant for their tasks related to recruitment and employee management process.
- LLM models
- RAG
- OpenAI Assistant
- GraphQL
-
2023.02 - 2024.08 AI Engineer
FPT Software
Developing and maintaining intelligent RAG LLM models. Developing a japanese license plate recognition and tracking system.
- LLM models
- Azure
- RAG
- Tracking
- Azure Services
- Kubernetes
- PaddleOCR
- Bytetrack
-
2023.01 - 2022.10 Machine Learning Engineer
Kyanon Digital
Developing a demo version of a pipeline for recognizing people, reID, pose detection, and tracking.
- Docker
- AWS
- Computer Vision
- ReID
- Pose Detection
- Tracking
- MySQL
Education
-
2019 - 2023 Ho Chi Minh City, Vietnam
Bachelor
University of Information Technology, Ho Chi Minh City, Vietnam
Data Science
- Machine Learning
- Deep Learning
- Computer Vision
- Natural Language Processing
- Data Structures and Algorithms
Publications
-
2022.11.03 A Novel License Plate Image Reconstruction System using Generative Adversarial Network
IEEE
This paper proposes a two-stage deep learning based algorithm for license plate reconstruction in parking lot management systems. The first stage uses YOLOv4 transfer learning to detect license plate positions, while the second stage employs Pix2Pix GAN to reconstruct poor quality, unrecognizable license plate images into perceptible ones. The system aims to enhance license plate image quality for both human and computer recognition.
-
2022.08.17 Intelligent Scene Text Recognition in Streaming Videos
Springer International Publishing
This research explores scene text recognition for extracting text from images and videos, which is increasingly important for visual media on social networks. We evaluated modern deep learning architectures like SAST, EAST, and Rosetta on the Vietnamese Vintext dataset by VinAI, comparing their real-time performance in frames per second.
Skills
| Data Science | |
| Machine Learning | |
| Deep Learning | |
| Computer Vision | |
| Large Language Models | |
| RAG | |
| Fine-tune LLM |
Languages
| Vietnamese | |
| Native |
| English | |
| IELTS 7.0 |
Interests
| AI | |
| LLM | |
| Computer Vision | |
| Machine Learning | |
| Deep Learning |
Projects
- 2024.08 - Now
Talent AI Assistant
Led the development of an advanced AI-powered chatbot assistant for an enterprise HR platform. Key achievements include: (1) Implementation of automated daily reporting system for recruiter workforce management, (2) Development of an HR knowledge base with 92% accuracy in query responses, (3) Creation of an intelligent onboarding system with contextual navigation assistance, (4) Design and deployment of an AI-driven job matching algorithm for optimal candidate placement.
- LLM
- RAG
- OpenAI Assistant
- GraphQL
- Function Calling
-
Shieve Data Platform
Led the development of core form recognition and data extraction capabilities for processing scanned technical drawings as Lead Data Engineer. Architected and implemented the solution utilizing Azure Cloud Services and Document Intelligence to enable automated document processing and analysis.
- Azure
- Azure Storage
- Azure Document Intelligence
- Azure Cognitive Services
- 2024.02 - 2024.08
Advanced Generative AI Innovation
Served dual roles as Business Analyst and AI Engineer across multiple client sub-projects focused on developing generative AI solutions. Key deliverables included fine-tuned LLM models capable of zero-shot question answering, an AI-powered physiognomy analysis assistant, and an automated system for staff ticket generation, etc.
- Azure OpenAI
- Azure Cognitive Services
- Elastic Search
- Kubernetes
- Llama-2
- Smart chunking
- 2023.06 - 2023.07
ChatGPT-based Banking Assistant
Developed and implemented an AI-powered banking assistant leveraging natural language processing to provide accurate responses based on internal documentation. Achieved 80% response accuracy while delivering the solution within an accelerated one-month development timeline.
- Azure OpenAI
- ChatGPT-3.5
- Cognitive Search
-
Japanese Intelligent Tracking System
Responsible for the AI Engine which recognizes and tracks Japanese license plates, achieving 92% accuracy in real-time tracking at a Japanese parking station using a limited training dataset of ~300 images from the client.
- Computer Vision
- PaddleOCR
- Bytetrack
- Real-time Tracking