Dylan Nguyen

Designed and developed an AI Agent system for real estate consultants and property lawyers in the UK, automating legal research and client interaction workflows. Built a scalable RAG pipeline using Langchain, integrating multi-role prompt design for consultants, lawyers, and end-users. Leveraged AWS Bedrock, Azure OpenAI, and Amazon S3 for model hosting, prompt orchestration, and secure storage.

Led a core AI team of 20 to build Alphii, a no-code platform for orchestrating collaborative AI agent teams. Designed and implemented a multi-agent architecture where agents operate as role-based collaborators to complete complex business tasks. Built an AI automation workflow platform engine that allows users to visually connect and configure agents into end-to-end processes.

Developed an intelligent voice-based medical assistant for a Japanese virtual clinic, providing patients with 24/7 health consultation through natural spoken interaction. Designed a tailored RAG pipeline adapted for spoken language, optimizing query understanding and answer generation from structured and unstructured medical sources. Utilized AWS Polly for high-quality voice output and AWS Transcr

Designed automation workflows for callbot and chatbot systems used in banking operations. Built logic-driven conversation trees for both voice and text interactions using custom scripting tools. Integrated voice processing tools with NLP components for smoother, intent-based dialogue.

Built a multi-role chatbot and content generation platform for product image description and retail marketing automation. Developed a modular RAG pipeline using Langchain, LlamaIndex, and DSPy for multimodal product search and captioning. Used GPT, Qwen, Mistral, and Qwen-VL to generate product descriptions and image-based tagging for e-commerce use cases.

Designed and implemented a virtual try-on system powered by Stable Diffusion for fashion e-commerce. Combined generative techniques with depth estimation and object detection for realistic try-on outputs. Applied Meta’s SAM model to achieve high-quality segmentation of clothing items and human figures.

Built a deep learning-based visual localization model for GNSS-denied UAV environments. Designed preprocessing methods for shadow removal and perspective handling. Proposed a new dataset captured from UAV flights in non-urban zones.

Researched object-specific generation using Stable Diffusion. Designed conditioning strategies to enhance relevance between text and image. Conducted internal evaluations for prompt-guided generation effectiveness.

Researched and benchmarked optimization algorithms for Video Synopsis. Provided appropriate algorithm evaluation metrics and applied SOTA Object Detection and Object Tracking methods. Evaluated algorithms on Viettel company’s video tracking dataset and researched new methods for Video Frame Interpolation.

Researched the Try-on algorithm with Diffusion Model and tested it with Virtual Try-On SOTA methods. Developed segmentation methods for the human body in preprocessing data and implemented human keypoint detection with Openpose for data preprocessing.

Developed a mobile-friendly object detection (Yolov5) model for road damage recognition. Handled data imbalance through augmentation and training optimization. Deployed the solution to Android devices using TensorFlow Lite.

Researched and wrote a paper on a new approach to stock chart pattern recognition. Built a dataset of stock chart patterns from historical data using MQL4. Created special data processing methods such as segmentation for candlestick charts and generated variation patterns to enrich the dataset.

Activities and societies: Bronze Medal of UW-Madison GI Tract Image Segmentation AI Competition Fall 2022; Published Paper Research at ICSCA/ACM Conference; Sprints 2023 Research Assistant at AI Vietnam; Leader of AI Research Team at FPT Innovation LAB Sum 2021; Research Assistant at Vietnam Institute of Military Science and Technology Sum 2023.

Completed a comprehensive program focusing on the principles and applications of artificial intelligence. Gained expertise in machine learning, deep learning, and data science techniques relevant to AI development.