Customer Relationship Management – USA
AI Engineer, 04/2025 - Now
Project description
- Working in an NLP project with 10 members including: 2 AI Engineers, 1 Front-end Developer, 2 Back-end Developers, 2 Business Analysts, 1 Quality Control, 2 Designers
- Create a web application that leverages OpenAI LLMs to automate sales deal assessments. It ingests structured deal and user profile data, applies deterministic scoring and probability forecasting, generates tailored, LLM-powered recommendations, and exports comprehensive reports. The platform also enforces input/output guardrails (including PII masking) for each of the components
Responsibilities
- Built an AI-powered sales intelligence platform that evaluates professionals and predicts deal outcomes
- Engineered a recommendation system that scores opportunities and generates actionable guidance
- Implemented comprehensive security measures with data protection and observability systems
- Designed a flexible reporting engine delivering customized outputs in multiple formats
Technologies
- Python, OpenAI, LangChain, Langfuse, Presidio Analyzer/Anonymizer
- PostgreSQL, SQLAlchemy, Alembic, Redis, Celery
- Docker, Docker Compose, Bitbucket Pipelines, pre-commit
Social & Dating Platform - USA
AI Engineer, 02/2025 - Now
Project description
- Working in an NLP project with 10 members including: 1 AI Engineers, 1 Front-end Developer, 1 Back-end Developers, 1 Business Analysts, 1 Quality Control
- Build a social datAI-driven search, matching and AI-driven search, matching and content-generation services over user profiles and preferences. It combines vector embeddings with a vector database to power lightning-fast similarity searches, multi-criteria partner matching and personalized recommendations. In addition to profile similarity and layered “visual” searches, the system offers batch compatibility scoring, AI-powered content moderation, generative hashtag creation and normalization, and a “Thought of the Day” personalization engine
Responsibilities
- Vector‐based similarity search
- Find “most similar” user profiles and partner preferences
- Support for exclusion filters, pagination and score ranking
- Layered / circle‐based searches, Overlap counting on combinable attributes
- Batch compatibility matching
- AI‐powered content moderation
- Hashtag generation & regeneration via LLM prompts
- Tag normalization to enforce consistency
- Personalized “Thought of the Day”: Insert, update and retrieve similar thoughts using embeddings + vector
Technologies
- Python, FastAPI, Pydantic, Qdrant, OpenAI API, LangChain, Langfuse
- Docker, Docker Compose, UV
Computer Vision for Healthcare – Singapore
AI Engineer, 08/2024 - Now
Project description
- Working in a Computer Vision project with 4 members including: 1 AI Engineer, 1 Python Dec
- Create a mobile app using OCR + LLM to scan and compare medication labels, ensuring prescriptions match dispensed medications for patient safety
Responsibilities
- Engineered and optimized an end-to-end OCR pipeline, integrating OpenCV-python for advanced image pre-processing (noise reduction, skew correction, layout analysis) to maximize recognition accuracy
- Deployed and managed machine learning models for OCR using onnxruntime, including selection, fine-tuning (if applicable), and performance monitoring of these models.
- Implemented solutions using Large Language Models (LLMs) for tasks such as named entity recognition, data categorization, summarization, and information extraction from unstructured medical text.
- Utilized Pillow for sophisticated image manipulation tasks integral to the OCR process.
- Applied geometric analysis libraries (shapely, pyclipper) for precise text region detection, segmentation, and extraction from complex document layouts.
- Employed tools for tracing, debugging, and evaluating the performance of LLM
Technologies
- Python, FastAPI, SQL Alchemy, Alembic, Celery, Redis
- ONNX Runtime, Paddle OCR, Pillow, OpenCV, Langchain, Langfuse
Healthcare – Singapore
AI Engineer, 05/2024 - Now
Project description
- Working in an NLP project with 4 members including: 1 AI Engineers, 1 Front-end Developer, 1 Back-end Developers, 1 Quality Control, 1 Designers
- A modular AI‐agent framework that provides a suite of LLM-powered assistants - each implemented as a configurable, graph-driven pipeline—that can be dynamically instantiated and orchestrated to perform specialized tasks such as health-package recommendation, document intelligence, customer support, and speech processing. The framework enforces shared guardrails for input/output validation, supports customizable LLM parameters per agent and is designed for high-throughput messaging via Redis/Celery within a FastAPI microservice.
Responsibilities
- Designed and implemented a Base Agent abstraction, agent constants, and a Factory pattern to dynamically create multiple specialized agents.
- Developed four domain-specific agents:
- A health-package recommender composed of nodes for age/gender extraction, question generation, info extraction, explanation and response mapping.
- A document-intelligence agent that extracts, parallel-processes, and summarizes clinical notes, IRS forms and questionnaires.
- A retrieval-augmented customer-support chatbot with human-escalation logic, knowledge retrieval, and context-aware response generation.
- An audio-processing agent offering speech ingestion, transcription and summarization pipelines.
- Implemented shared input/output guardrail nodes to validate prompts and enforce response schemas.
- Configured per-agent LLM parameters (chat history, embeddings, temperature, token limits) via a centralized params module.
- Integrated OpenAI API and LangChain for prompt construction, callbacks, embeddings and parallel execution.
Technologies
- Python, FastAPI, Pydantic, OpenAI API, LangChain, LangGraph, Lang
- Celery, Redis, PostgreSQL, Docker & Docker Compose
LEGAL PROFESSION– Internal usage
AI Engineer, 10/2023 - 05/2024
Project description
- Working together in an AI Team which has 5 members
- Create a GPT powered Chatbot for Knowledge base Q&A from uploaded documents in two formats: PDF and Docx
Responsibilities
- Build base code for the chatbot with multiple documents
- Create UI for demo application
- Create system design diagram
Technologies
- Python, Langchain framework, OpenAI API
HUMAN RESOURCES – Internal usage
AI Engineer, 10/2023 - Now
Project description
- Working together in an AI Team which has 5 members
- Create a GPT powered Chatbot for Knowledge Q&A and suggest potential candidates from database
Responsibilities
- Designing prompts and agents to use database tools
- Create UI for demo application
Technologies
- Python, Langchain framework, OpenAI API
TECHNOLOGY – Internal usage
AI Engineer, 12/2022 - 06/2023
Project description
- Working together in an AI Team which has around 12 AI Engineers
- Build back-end for a Chat-GPT like application for around 200 internal employees
Responsibilities
- Research and experience with summarizing models to integrate in the system
- Support in design database schema and project structure
- Fix bugs and maintain the source code along with supporting other team members
Technologies
- Python, FastAPI framework, PostgreSQL
- Trafilatura for gathering text on the Web
- Visual Studio Code, GitHub, Docker
LOGISTICS – Singapore
AI Engineer, 03/2022 - 09/2022
Project description
- Working on a Proof-Of-Concept project with 4 regular members and 2 temporary members including: 1 Project Owner/ Scrum Master, 1 Team Leader, 1 Consultant, 2 AI Engineers and 1 Label Engineer
- Prove that we can apply the Computer Vision technology in detecting damages on container using YOLO Framework
Responsibilities
- Enhance processing time of over 300 high-resolution images with various pre-processing steps from 4 hours to less than 5 minutes
- Train, evaluate and test over 20 object detection models using YOLOv5 for 3 different types of objects
- Support in creating labelling guidelines for consistent data quality
Technologies
- Google Cloud Platform, Vertex AI
- Python, YOLOv5 Object Detection Models
- Label Studio, LabelImg, labelme tools
- Visual Studio Code, Pycharm, GitHub
ECONOMICS – University Project
Junior Data Science, 10/2021 - 1/2022
Project description
- Working together in a team of 2 people
- Analyze and forecast for US Crude Oil Production from May 2021 until December 2021 using different forecasting methods
Responsibilities
- Performing data analysis on time series data to discover insights and patterns
- Build a total of 6 models for forecasting the Crude Oil Index using: Random Walk with Drift for benchmarking, (Error, Trend, Seasonal) model, STL decomposition with ETS approach, Seasonal ARIMA, Regression with ARIMA errors and Dynamic regression with Fourier terms
Technologies
- Visual Studio Code, Pycharm, GitHub
