MarketEnterprise Vietnam
AI & System Optimization Engineer / Full-Stack Engineer
Participated in the development team redesigning Oikura's architecture — a second-hand goods trading platform in Japan — from monolithic to microservices to optimize performance.
- Applied NX Monorepo, established clear boundaries between internal and public services
- Developed NestJS microservices following SOLID principles, optimized MySQL queries
- Contributed to building automated CI/CD pipelines on AWS (EC2 + CodePipeline)
- Integrated notification systems (SendGrid, SMS) through event-driven architecture with SQS & Lambda
Contributed to developing an automated voicebot system for information collection at takakuureru.com service, using AWS services.
- Designed state machine and conversation flow for automated data collection
- Integrated Bedrock Claude Haiku for intent classification and structured data extraction
- Handled real-world voice communication issues: dead air, brand recognition errors, PostCode formatting
- Set up protection mechanisms (rate limit, interaction limits) to protect the system
Designed and deployed an automated pipeline for transcription, summarization, and data extraction from IS/FS business calls.
- Built end-to-end processing flow: STT → Diarization → Summarization → Structured Extraction
- Used SageMaker async endpoints and Lambda workers to optimize cost and latency
- Fine-tuned for improved accuracy on real-world audio data (background noise, cross-talk)
- Stored and exported data via S3 Tables and Google Sheets for operations team
Experimented and deployed a multi-agent system to support and partially automate IS/FS staff workflows.
- Designed multi-agent workflow using LangGraph and Gemini
- Applied Agent-to-Agent (A2A) communication standards to synchronize data between Voicebot and decision-making system
- Optimized memory and caching mechanisms for agents
- Flexibly integrated with existing infrastructure (S3, DynamoDB)
Participated in developing an event-driven customer communication management system for internal operations.
- Used AWS SAM, Lambda, and EventBridge for data synchronization and task scheduling
- Built backend workflow with NestJS and MongoDB database
- Created ReactJS interface for operations staff, integrated real-time webhooks (SMS, Gmail, Slack, Line)
Researched and developed a chatbot tool to support retrieval and analysis of internal technical documents.
- Improved search result accuracy using Word2Vec embeddings
- Deployed LLM model (Ollama) on internal network for information security
- Designed modular pipeline, easy to extend and optimize response time