VNG
AI tóm tắt
Skills bắt buộc
About the role & product We are building an **Agentic Platform** — a platform that helps customers deploy and operate AI agents at production scale, quickly and safely. The platform provides ready-made services / components so agent builders can integrate quickly instead of building from scratch — significantly shortening the time from idea to a finished agent. The platform also follows open protocols to easily connect with tools, data sources and other agents. We believe a solid agentic platform is built on **foundational software engineering (backend & distributed systems)**, with AI / agents as the capability layer on top. Therefore, this role is designed with a **70% traditional software engineering / platform engineering** and **30% AI, AI agent and related protocols** split. You will work directly with the engineering team to take the services powering the platform from architecture, implementation, to operation — ensuring high availability, low latency, security and scalability.
Main responsibilities 1. Platform & Backend Engineering (~70%) - Design, implement and maintain backend services / APIs that meet high standards of **availability, latency and security**. - Build the platform's infrastructure components to support the agent lifecycle — from where agents run, communication between agent / tool / service, state management, to security and authorization. - Build a **secure sandbox runtime** where AI agents can execute generated code and tools in isolation — using containerization / microVMs (Docker, gVisor, Firecracker), namespace & seccomp isolation, resource limits (CPU / memory / network), and strict egress controls so untrusted agent actions never leak into the host or other tenants. - Design data models and data flows on SQL (MySQL), NoSQL (MongoDB) and vector databases; ensure consistency, throughput and recoverability. - Integrate message brokers / event-driven backbones (Kafka, RabbitMQ, AWS SQS/SNS) for asynchronous communication between the platform's internal services. - Identify and resolve system issues: performance bottlenecks, memory leaks, race conditions, security vulnerabilities, resource leaks. - Design and deploy **observability** systems (metrics, logs, distributed tracing) to monitor and debug the platform in production. - Research, experiment with and evaluate new technology solutions to solve the system's technical problems. 2. AI & Agentic Capability (~30%) - Design foundational services for AI agents; proactively keep up with trending AI agent features in the market so customers always access the latest capabilities. - Build and integrate **MCP servers** as well as adapters for the **A2A** protocol to connect agents with tools, data sources and other agents. - Integrate LLMs / agentic frameworks (LangChain, LangGraph, CrewAI, Strands…) into the platform in a framework-agnostic way; design abstractions so the platform does not depend on any specific model or framework. - Design and build the **agent sandbox capability** — a code / tool execution environment (code interpreter, REPL, tool-call runtime) that gives AI agents the ability to run code, evaluate outputs and iterate safely; support multi-language execution, persistable session state, file I/O, and deterministic reproducibility so agents can reliably "think → execute → observe → refine". - Define and enforce **sandbox security boundaries** — permission scoping per agent / per task (read-only fs, network allow-lists, approval gates for sensitive actions), and deep integration with the platform's authorization layer so sandbox actions are auditable, traceable and revocable. - Write clients (SDK, CLI, AI agent skills) to help customers integrate and use the platform. - Apply AI tooling (coding assistant, AI code review, AI-assisted testing & debugging) into the daily development and operations workflow to boost team productivity.
Job requirements I/ Must-have 1. Background & experience - Bachelor's degree or above in Computer Science, Engineering or a related field. - Experience building backend services / distributed systems running in production; for us, years of experience is only a reference number — **actual capability** is the deciding factor. - Solid foundation in **software architecture, design patterns, distributed systems** and best practices. - Strong problem-solving, logical thinking and analytical skills; effective communication and collaboration.
II/ Nice-to-have - Have built or contributed to a real **agentic platform / AI platform**. - Have written an **MCP server** and integrated it into a production system. - Have deployed an end-to-end **monitoring / observability** system. - Experience with **Kubernetes** and operating container workloads at scale. - Experience with **code interpreter sandbox**, identity / authorization for agents, or semantic caching. - Have built a **secure sandbox / code execution runtime** for AI agents or multi-tenant workloads — hands-on with container isolation (Docker, gVisor, Firecracker / microVMs), namespace & seccomp, network policies, or serverless code execution platforms. - **Good English listening and speaking skills** — a plus for working with international teams, partners and technical documentation.
Nguồn: careers VNG
