Tuan Anh
Nguyen

• Designed and developed scalable Camera AI & Video Management Systems (VMS) for real-time video analytics, scaling from 200 to thousands of cameras operating 24/7 at 25 FPS with >90% accuracy.
• Deployed and optimized production AI workloads across multiple NVIDIA GPU types (RTX 4090, RTX 5090, RTX PRO 6000) using CUDA, NVIDIA drivers, Docker, and Kubernetes.
• Optimized YOLO-based real-time object detection pipelines using NVIDIA TensorRT, FP16 precision, and GPU acceleration to improve inference speed, reduce latency, and enhance GPU resource utilization for scalable computer vision applications.
• Trained, fine-tuned, and deployed computer vision and multimodal models using YOLO, YOLO Pose, VLMs, LLMs, and MediaPipe, including:
  • Fight Detection VLM: fine-tuned Qwen3-VL-4B with 4-bit QLoRA (0.74% of weights trained). Cut errors from 38 to 4 on 322 held-out images; accuracy 88.2% → 98.8% across 5 seeds.
  • Built a perceptual-hash audit to catch train/test leakage and fixed an image-resolution bug, cutting visual tokens from ~1,450 to ≤390 per image; ~110 ms p95 latency (offline benchmark, single RTX 4090).
  • Speech Moderation LLM: fine-tuned Qwen2.5-1.5B with QLoRA (3.4 GB VRAM) to output structured JSON verdicts, severity levels, and flagged terms. Beat zero-shot, lexicon, and TF-IDF baselines: 2× term-level F1 vs. zero-shot; invalid JSON 28 → 1.
  • Found and fixed a dataset-parsing bug affecting ~4,000 training records; raised throughput 67× (4.4 → 294 req/s, 111 ms p50) via LoRA merging and vLLM (offline benchmark, RTX 4090).
• Built and maintained backend services and RESTful APIs in Python/FastAPI for production AI applications.
• Developed frontend and desktop applications with React.js and Electron, integrating AI features into user-facing products.
• Owned the full AI lifecycle: training, fine-tuning, inference optimization, backend, frontend integration, and infrastructure deployment.

• Designed and developed an AI-powered travel chatbot for a platform with 100k–200k users, understanding travel intent and recommending POIs, hotels, and travel experiences in context.
• Built and optimized AI workflows integrating LLMs, travel APIs, search services, and contextual trip data; integrated Google Gemini with Serper (web search) and Nuitee (hotel and travel data).
• Researched and developed facial recognition applications for real-world AI systems.
• Developed and maintained backend services with FastAPI and NestJS for scalable AI and travel workflows.
• Designed and ran unit, integration, and stress tests to validate chatbot reliability and behavior under high load.
• Implemented production monitoring and AI observability with Langfuse (LLM traces, latency, token usage, pipeline performance).
• Containerized and deployed production services with Docker and Fly.io; worked with product and engineering teams on production issues and UX optimization.
• Worked with Fly.io platform infrastructure, covering core server, container, storage, and networking services.

• Led architecture and development of AI-powered sales and marketing products, owning the stack across frontend, backend, AI systems, and DevOps.
• Built automated marketing systems with scalable pipelines and workflow automation.
• Developed AI meeting agents that join meetings, capture recordings and transcripts, and use OpenAI, Claude, and Llama to generate insights, analysis, and marketing strategies.
• Designed and implemented RAG systems over user-specific knowledge for contextual analysis and recommendations.
• Built AI voice agents with real-time voice via Twilio and ElevenLabs.
• Designed scalable AI services and production infrastructure using microservices, containerization, and cloud-native technologies.
• Worked with AWS infrastructure: EC2 (servers), ECS and ECR (containers), S3 and EBS (storage), and VPC, Route 53, and Application Load Balancer (networking).

• Completed 26+ projects and 1,100+ hours for international clients with a 100% Job Success Score and 4.9/5 average rating.
• Delivered production-grade LLM applications, RAG systems, computer vision, voice AI, automation platforms, and scalable backend infrastructure across multiple industries.
• Worked directly with founders, CTOs, product and engineering teams to turn business requirements into technical solutions, owning projects from architecture through deployment and production support.
• Solved engineering problems in system architecture, AI performance, GPU optimization, scalability, and production reliability.
• Worked with AWS infrastructure for client projects: EC2 (servers), ECS and ECR (containers), S3 and EBS (storage), and VPC, Route 53, and Application Load Balancer (networking).
• Worked with PostgreSQL relational databases for client backend services alongside MySQL and MongoDB.

Key client: Lighthouse Education (United States) — PHP Developer · Mar 2025 – Sep 2026

• Developed and deployed web applications using PHP and CodeIgniter.
• Built AI-powered educational chatbots for interactive learning, integrating Google Gemini APIs.
• Developed backend services in PHP and Python FastAPI, integrating LLM capabilities into production; data and knowledge infrastructure on MySQL, MongoDB, and Weaviate.

• Designed and developed scalable API systems with Python FastAPI and Node.js; designed resilient infrastructure on Kubernetes and Docker.
• Managed project workflows, assigned tasks, tracked KPIs, and guided engineering teams; defined product strategy and technical roadmaps with company leadership.
• Researched and implemented database sharding for large-scale data processing.
• Developed real-time communication systems with MQTT, WebSocket, and WebRTC for chat, VMS, and attendance applications.
• Led AI/ML R&D across object detection, facial recognition, vectorization, OCR, audio processing, and LLM-based applications; built vector search on Elasticsearch and Pinecone.
• Built and maintained web and mobile applications with React and Flutter; supported automated data crawling systems.

• Managed and supervised engineering tasks for an app with ~1M registered users, providing technical support and resolving complex implementation issues.
• Developed a real-time video calling application with WebRTC, MQTT, and WebSockets, and desktop applications with Electron.js.
• Led migration of legacy systems from PHP/MySQL to Node.js/MongoDB for scalability and maintainability.
• Built and deployed Next.js applications focused on performance and SEO: ~200 ms page load, Google PageSpeed >90.
• Managed production infrastructure (Nginx, PM2, Docker, firewall, domains); applied distributed storage and system-level optimizations; supported AI-driven applications.