Luong Quang Dung
AI Engineer
I'm an AI Engineer with a solid track record of turning advanced AI models into real-world, production-ready systems. I began my career at FPT Software, developing scalable real-time video analytics and OCR systems for edge devices and servers using DeepStream, Triton, and microservices. There, I gained deep expertise in optimizing performance, accuracy, and deployment pipelines for AI solutions.
I now work at Success Software Services, focusing on large language models and conversational AI. My recent work includes building a Retrieval-Augmented Generation (RAG) platform with high-performance vector search and a real-time Voice Agent for healthcare and wellness using WebRTC and streaming inference. I’m passionate about combining high-performance engineering with natural, human-like AI experiences, following best practices in MLOps, observability, and scalability.
Experience
AI Engineer
May 2025 - PresentSuccess Software Services
- •Designed and deployed a RAG platform with FastAPI, vLLM, and Milvus to deliver precise, context-aware answers.
- •Created data pipelines for PDF ingestion, embeddings, and vector indexing with Airflow and Docker-based microservices.
- •Architected a real-time Voice Agent for healthcare and wellness using WebRTC, integrating STT, LLM reasoning, and TTS modules for natural conversations.
- •Implemented scalable microservice architecture with streaming inference, quantization, and observability to improve speed and reliability.
AI Engineer
June 2022 - May 2025FPT Software Quy Nhon
- •Built and optimized real-time video analytics systems for Jetson edge devices and servers using NVIDIA DeepStream, Triton, Redis, Kafka, and Docker.
- •Delivered LLM-based chatbots and context-extraction systems integrating semantic retrieval and contextual memory.
- •Implemented OCR and object detection pipelines achieving up to 98% accuracy, improving efficiency and inference time.
- •Conducted performance optimizations and model fine-tuning to improve system accuracy and efficiency.
- •Collaborated with cross-functional teams to deliver high-quality solutions within tight deadlines.
Projects
End-to-End RAG Platform
An AI operations platform for managing LLM deployments with advanced RAG capabilities.
Codewiki
Codewiki is an intelligent documentation tool that automatically generates wiki-style documentation for your GitHub or local repositories using AI.
DeepStream YOLOv11
Plug-and-Play Custom Parsers for AI Models in NVIDIA DeepStream SDK. Supported YOLOv11, D-FINE, SCRFD model.
Face Recognition System
Face recognition pipeline powered by Triton Inference Server.
Awards & Recognition
Best Performer of FPT Software 2024
January 2025FPT Software Quy Nhon
Recognized for dedication and contributions to team success
1st runner-up MLOps Marathon
September 2023Vietnam MLOps Marathon
A competition focused on real-world business challenges, and honing AI project deployment skills
Winner HuggingFace Competition
April 2023HuggingFace
Detecting Ships in Ports to Avoid Congestion and Manage Traffic
Top 4 Final Quy Nhon AI Hackathon
September 2022FPT Software
Placed second in Qualifying, advanced to the Finals, and finished in the top 4