Ho Chi Minh City, Vietnam
Đỗ Xuân Lộc
AI & System Optimization Engineer
Full-Stack Engineer · Solutions Architect (ML)
From Full-Stack to AI/System Optimization — the journey of building real-world, scalable, cost-optimized systems for the Japanese market.
Node.jsNestJSAWS (EC2, ECS, Lambda, SAM, SQS, SNS, EventBridge, ECR, CodePipeline, SageMaker, Bedrock)RAGWord2Vec / Embeddings +41 skills
- 01Systems built to outlast their architects.
- 02Intelligence woven in — not bolted on.
- 03Infrastructure that earns back its cost.
Problems I demolish
What I'm built for
Real challenges, real approaches — each card a battle I've fought and systematized.
01
Large-scale systems difficult to scale and maintain
Decompose into microservices within NX Monorepo, clear API boundaries, automated CI/CD pipelines per environment.
NestJSNX MonorepoAWSCI/CD
02
Synchronizing and processing multi-source realtime data
Event-driven serverless architecture — queues and schedules decouple services for easy scaling with load.
AWS LambdaEventBridgeSAMSQS
03
Integrating LLMs into products while maintaining accuracy and safety
RAG anchors answers to real documents, embeddings + internal LLM deployment to control data and reduce hallucination.
RAGEmbeddingsLLM (Ollama)Python
04
Handling real-world Japanese voice interactions with high accuracy
Combine Amazon Connect + Lex V2 + Bedrock Claude Haiku; handle dead air, ASR brand errors, and complex flows.
Amazon ConnectLex V2BedrockState Machine
Tech DNA
Full-stack to AI — the complete stack
Every layer from UI to inference pipeline, hardened through production.
Frontend 8
ReactJSNext.jsReduxTailwindCSSVueJSVuexVuetifyTypeScript
Backend 6
Node.jsNestJSExpressSocket.ioSSEThree.js
Cloud/DevOps 6
AWS (EC2, ECS, Lambda, SAM, SQS, SNS, EventBridge, ECR, CodePipeline, SageMaker, Bedrock)DockerGitHub ActionsNX MonorepoCI/CDCloudFormation
AI / ML 9
RAGWord2Vec / EmbeddingsLLM (Ollama, Claude, Gemini)Fine-tuning (LoRA/QLoRA)Whisper + PyannoteFlorence-2LangGraphMulti-Agent SystemsAI Pipelines
Practices 7
MicroservicesEvent-Driven ArchitectureDevSecOpsScrum/AgileMonitoring & ObservabilityBusiness IntelligenceCost Optimization
41 skills across 5 domains
Trajectory // Roadmap
Future roadmap
From today's AI core to a far horizon — where I'm headed.
01
2026 In progress
AI Core Forge Forging real-world AI core — turning every system into an intelligent entity.
Production RAGLLM optimizationObservabilityInference tuning
02
2026–2027 Planned
Mind at Scale Scaling artificial intelligence for large-scale use without sacrificing performance.
Distributed inferenceMLOpsServerless AIAuto-scaling
03
2027–2028 Planned
DevSecOps Ascent Armoring AI with steel — security and cost optimization from the inside out.
Zero-trust AISecure CI/CDCloud cost controlCompliance
04
2028–2030 Planned
AI Solutions Lead Leading large-scale AI architecture and shaping solutions for global enterprises.
Enterprise RAGAgent orchestrationPlatform architectureAI governance
05
2030+ Planned
N3 Global Bridge Connecting Vietnam with Japan and the world through AI — powered by N3+ Japanese proficiency.
Japanese N3+Cross-border teamsAI ethicsThought leadership
Blog
Latest posts
2026-06-06 · 6 min read Lambda Managed Instances: khi serverless cần thêm sức mạnh và sự linh hoạt của EC2 AWS mở rộng Lambda để chạy trên EC2 instances do họ quản lý. Đây là bước tiến thú vị cho những ai đang tối ưu hệ thống event-driven trên cloud, nơi pure serverless không còn là câu trả lời duy nhất cho mọi workload. 2026-06-05 · 4 min read Xây portfolio tự cập nhật hàng ngày bằng Grok × Claude Tại sao và làm thế nào mình dựng một portfolio tự sinh tin tức & blog mỗi ngày, với Grok làm content-engine và Claude làm architect.
Tech radar