Software Engineer, EYC
Nguyễn Ngọc Phúc Duy is a software engineer at EYC, working on building and running the company's software systems.
How to integrate Zalo OA with your CRM and an AI agent: OA OpenAPI vs ZBS Template Message (formerly ZNS) vs Mini App, webhooks, token handling, architecture, user consent and a rollout checklist.
RAG (retrieval-augmented generation) lets a language model look up your own documents before answering and cite its sources. This guide covers how it works, Vietnamese-language pitfalls, permissions, evaluation and a rollout plan.
MCP is an open standard that lets AI applications connect to external data and tools through one shared protocol. This guide explains the architecture, security, how it compares with APIs and function calling, and when a business should adopt it.
Prompt injection is an attack that uses text to make a language model follow an attacker's instructions. This guide explains direct and indirect injection, why tool-using AI Agents raise the stakes, and how to defend in layers using OWASP and NIST guidance.
What preschool parents actually need from a communication app, how class chat groups compare with a dedicated app, and how to handle photos, children's data, notifications and grandparents.
A preschool classroom TV is most useful as a display for the daily schedule, visual cues and announcements — not for children to sit and watch. How to use it, what WHO and AAP recommend, placement, privacy and device management.
A practical guide to smart displays and IoT in business operations: common use cases, device–gateway–cloud architecture, device management and OTA updates, security baselines from NIST, ETSI and OWASP, offline behaviour and privacy.
A practical cloud cost optimization guide based on the FinOps Framework and the AWS, Google Cloud and Azure cost guidance: 12 concrete actions from tagging and budgets to rightsizing and a monthly review.
A practical guide to choosing an LLM for Vietnamese using verifiable criteria: language quality, domain, latency, cost drivers, data residency, licensing, tool use and context length, plus how to build your own evaluation set.
How to pick the right process, map the workflow, place human approval, run a measured pilot and only then scale — with a scoring table and a list of what not to automate.
Edge AI means running AI models directly on devices or on-site servers, close to where data is created. This guide compares edge, cloud and hybrid setups and ends with a decision checklist.