Hi, I’m Cuong

I build full-stack AI apps, from agentic RAG systems to scam-detection tools, serving thousands of real users.

AI & Full-stack Web

Tech Lead atChongLuaDao

Ho Chi Minh City, Vietnam

Le Nguyen Viet Cuong

Experience

Tech LeadatChongLuaDao

Jan 2025Present

Key contributions

  • Led a team of 4 to build an AI scam website detector with 300+ daily active users, earning 2 national awards: Nhan Tai Dat Viet and Make in Vietnam.
  • Built an agentic RAG scam adviser with vector-based search, keyword search, tool calling, and real-time web context for 3,000+ users.
  • Fine-tuned an open-source LLM using LoRA to balance cost and performance.
  • Scaled legacy APIs to millions of daily requests with monitoring and alerts.
  • Technical lead for 10+ international partners.

Tech stack

  • Next.jsNext.js
  • SolidJSSolidJS
  • TypeScriptTypeScript
  • PythonPython
  • PostgreSQLPostgreSQL
  • RedisRedis
  • MinIOMinIO
  • DockerDocker

Reference letter

From Ngo Minh Hieu (Hieu PC)CEO of ChongLuaDao

Reference letter from Ngo Minh Hieu (Hieu PC)

Products shipped

ai.chongluadao.vn
ai.chongluadao.vn

Scam website detector powered by LLMs and threat intelligence, with a live browser agent.

zalo.me/chongluadao
zalo.me/chongluadao

Chatbot on Zalo that checks for scam signals and answers online safety questions based on ChongLuaDao's knowledge base.

tracnghiem.chongluadao.vn

Game about spotting real vs. fake websites, emails, messages, videos, and images.

baovethongtin.com

Web app that improves users' digital security via personalized checklists, assessments, and progress tracking.

Full-stack Developer & AI IntegratoratLumist

MayAug 2025

Key contributions

  • Built a science-backed AI vocab system used by 1,000+ learners.
  • Added personalized LLM-powered test analysis and student question answering.
  • Resolved a critical bottleneck to deliver 5x faster page loads.

Tech stack

  • Next.jsNext.js
  • TypeScriptTypeScript
  • SupabaseSupabase
  • Vercel AI SDKVercel AI SDK

Product shipped

lumist.ai
lumist.ai

An all-in-one Digital SAT prep platform with an extensive question bank, AI-driven performance analysis, predictive scoring, gamification, and more.

Projects

Examinai

AI-powered IELTS Learning Platform

Writing feedback breakdown
Grading dashboard
Speaking assessment view
  • Integrated multimodal AI for text generation, image understanding, text-to-speech, and speech-to-text.
  • Built reliable writing and speaking assessment flows based on official rubrics and multiple LLM judges.
  • Built an agentic RAG question-answering system grounded in an IELTS knowledge base.
  • Curated 1,000+ practice tests via web scraping.

Tech stack

  • Next.jsNext.js
  • TypeScriptTypeScript
  • PythonPython
  • SupabaseSupabase
  • Vercel AI SDKVercel AI SDK

InfiniteCompute

E-commerce Store for GPUs with AI Advisor

AI advisor recommending a build
GPU catalog
Admin analytics dashboard
  • Built a full e-commerce store with checkout, reviews, coupons, and admin tools.
  • Created an agentic RAG recommendation system combining product database search, internal documents retrieval, and real-time web context.
  • Dockerized all services to support reproducibility across environments.

Tech stack

  • Vue.jsVue.js
  • PythonPython
  • FastAPIFastAPI
  • PostgreSQLPostgreSQL
  • DockerDocker
  • OpenRouterOpenRouter

Research

Published

A Novel Spam Review Detection Architecture Using Large Language Models

Spam review detection architecture: ViSpamReviewV2 dataset feeding a preprocessing pipeline into the Llama 3.1 model
  • Fine-tuned 3 LLMs using LoRA to detect spam reviews on Vietnamese e-commerce platforms.
  • Improved accuracy of prior work from 88.54% to 92.12%, and F1 score from 73.49% to 78.86%.
  • Deployed scalable, cost-efficient serverless GPU inference for the models.
Read the paper
Under Review

Beyond URLs: Content-Aware Malicious Website Detection Using Multilingual BERT

  • Trained 20+ models (transformers, traditional ML models, hybrid ensembles) in parallel using cloud GPUs, reaching 99.06% accuracy on a 68,954-sample test set.
  • Surpassed 4 prior works by up to 8.28% in accuracy.
  • Compiled and cleaned a large real-world dataset of 689,556 webpages.
  • Designed a preprocessing pipeline to cut input noise by over 96%.

Recognition

Vietnam Television (VTV) coverage

VTV9 nationwide anti-scam panel
Dec 2025

VTV9 nationwide anti-scam panel

Panelist on VTV9's nationwide anti-scam campaign with Viettel Security and the Ministry of Public Security, sharing the tech perspective alongside policy and law-enforcement experts.

VTV9 talk show: AI-powered scam prevention
Sep 2025

VTV9 talk show: AI-powered scam prevention

Featured guest on VTV9's Kính Đa Chiều talk show, breaking down how AI can be used to detect and stop online scams targeting Vietnamese users.

VTV9 talk show: Living in the age of AI
Dec 2025

VTV9 talk show: Living in the age of AI

Returning guest on Kính Đa Chiều to discuss the broader impact of the AI era, from everyday productivity to the new risks it brings.

VTV9 talk show: Inside AI-powered scam tactics
Mar 2026

VTV9 talk show: Inside AI-powered scam tactics

Invited back to Kính Đa Chiều to walk viewers through emerging AI-driven scam techniques like deepfakes, voice cloning, and AI-generated phishing.