Insurance - Australia
AI Developer, Apr 2023 - Present
Project description
- Led the development of a virtual assistant designed to support disabled individuals in accessing Social Security Disability Insurance. The assistant was architected with a RAG-based framework, integrating OpenAI APIs for intelligent
Responsibilities
- Identify the specific needs of disabled individuals, determine the functionalities required.
- Integrate OpenAI API for natural language processing and understanding.
- Gather relevant data, using RAG to enhance the assistant's ability to provide accurate information.
- Design effective prompts to guide the assistant's responses.
- Implement function calls to handle specific tasks.
- Deploy the assistant on Azure.
- Monitor its performance and make necessary adjustments
Technologies
- Python, Pytorch, Oracle, MySQL, Azure, Networking.
Mobile Test Automation Platform - US
AI Developer, Feb 2021 – Apr 2023
Project description
- Engineered a computer vision–powered automation pipeline to validate UI/UX in mission-critical mobile delivery workflows.
Used deep learning to enable visual understanding of mobile app interfaces.
Responsibilities
- Construct a pipeline based on Mask-RCNN from Detectron2 to detect UI elements in mobile applications (semantic segmentation).
- Build a model to detect target elements from overlapping UI elements on a user touchpoint (box classification).
- Implement a panorama stitching algorithm using OpenCV to stitch multiple overlapping UI screens (image stitching).
- Implement template matching using OpenCV to match UI elements (template matching)
Technologies
- Python, Pytorch, Triton, AWS, Fast API, Flask, PostgreSQL, Docker, Kubernetes, Redis.
E-commerce - Vietnam
AI Developer, Mar 2019 – Feb 2021
Project description
- Built a suite of AI systems for one of Vietnam’s largest e-commerce platforms to manage inventory risk, verify seller legality, and ensure image moderation.
Responsibilities
- Construct a logo recognition pipeline based on Yolov3 for the logo localization step and ResNet for the logo classification step to detect OEM/non-OEM products to help validate millions of sellers' products (logo recognition).
- Construct an end-to-end system based on Tesseract and image preprocessing using OpenCV to extract the business registration code from the business registration certificate and check a seller's legality (BRC recognition).
- Develop an API based on the ResNet model to detect NSFW images and verify all photos uploaded on Tiki (NSFW recognition).
Technologies
- Python, AWS, Pytorch, OpenCV
Conversational AI Platform - Singapore
Machine Learning Engineer, Jul 2016 – Apr 2018
Project description
- Created intelligent recommendation engines for a recruiting startup, matching candidates with jobs based on skill semantics.
Responsibilities
- Built NLP pipeline to extract skills and job roles from CVs and job descriptions using Spacy + FastText.
- Used gradient boosted trees and cosine similarity metrics for personalized job ranking.
- Built an internal tool to auto-summarize CVs using extractive text summarization.
Technologies
- Python, Scikit-learn, XGBoost, FastText, Flask
