AI Architect (內湖瑞光)

We are looking for an AI Architect to own the technical architecture of Advantech’s enterprise AI platform — the shared “AI Factory” that we builds once and reuses across every business domain. The Architect owns the platform architecture, technical standards, and platform decisions that make AI agents and applications practical, secure, scalable, and cost-effective.

This is a hands-on, senior individual-contributor role. You will set the architecture across models, agents, data, and integration; codify patterns and guardrails that engineers and domain teams reuse; and stay close enough to the code to prototype, review designs, and unblock delivery.

• Platform architecture — Define and own the end-to-end platform architecture for enterprise AI agents and applications across the model, skill/MCP, data, and integration layers, and keep it coherent as the platform grows.
• Model platform — Architect the NemoClaw model platform — model evaluation and benchmarking, fine-tuning (SFT / LoRA / RLHF), model gateway and routing, harness engineering, and inference optimization.
• Agent & skill ecosystem — Design the skill/MCP platform — enterprise skill registry, MCP server framework, and agent orchestration and A2A standards — with versioning and reuse governance so one build serves many domains.
• Data & RAG architecture — Establish the data platform architecture: semantic layer, master data, vector stores, knowledge bases, and RAG pipelines that keep agents grounded in accurate, well-governed enterprise data.
• Security, guardrails & compliance — Set the standards for on-premises, data-isolated deployment — RBAC and audit trails, prompt-injection defense, guardrail agents, and responsible-AI review gates — and make them the default in every solution.
• System integration — Design robust, secure integration patterns so agents can read and write real enterprise systems through well-governed APIs, data, and permissions.
• Agent DevOps & observability — Define agent CI/CD, distributed tracing and telemetry, eval-driven regression testing, and token-cost and capacity practices that keep agents reliable, observable, and cost-aware in production.
• Technical standards & enablement — Codify architecture patterns, reusable implementation templates, and guardrails that AI engineers and domain AI Business Partners can reuse; raise the team’s technical bar through design reviews and mentoring.
• Frontier technology evaluation — Continuously scan, benchmark, and controlled-adopt frontier models, agent frameworks, RAG stacks, and emerging protocols; make build-vs-buy and platform-direction recommendations.
• Hands-on delivery partnership — Stay hands-on — prototype, review code and designs, and work side by side with AI engineers and the AI Project Manager to turn architecture into shipped, adopted solutions.
• Scalability & cost — Ensure the platform scales across domains within clear performance, reliability, and cost targets, and evolve it as adoption and usage grow.
• Broader IT architecture — Provide architectural guidance to other enterprise application and system-integration projects when needed.

• 7–12 years of experience in software engineering, AI/ML or data engineering, or solution architecture, including 3+ years designing production AI/ML or large-scale distributed systems.
• A proven track record of designing and shipping production systems built on LLMs and agentic architectures (RAG, tool use, orchestration) — not only prototypes or proofs of concept.
• Deep, hands-on expertise across the modern agent stack: models and inference, prompt and harness engineering, vector stores and retrieval, and APIs and system integration.
• Strong software-engineering fundamentals, with the ability to set and enforce architectural standards, run design and code reviews, and mentor engineers.
• Experience with cloud AI platforms (Azure and/or AWS) and with on-premises or data-isolated deployment.
• A solid grasp of AI and data governance, security controls, and human-in-the-loop design.
• Ability to translate business needs into architecture and to communicate technical direction clearly to both engineers and business stakeholders.
• Bachelor’s or Master’s degree in Computer Science, Engineering, or a related field, or equivalent practical experience.
• Professional working proficiency in English and Mandarin Chinese.

• Experience building an internal AI platform — a model gateway, an MCP/skill registry, or an agent orchestration framework.
• Familiarity with fine-tuning techniques (SFT, LoRA, RLHF) and with model evaluation and benchmarking.
• Experience with observability and eval-driven testing for LLM/agent systems, and with token-cost and capacity management.
• Knowledge of enterprise systems and integration (CRM/SFDC, e-commerce, quotation, ERP) and API-gateway design.
• Exposure to the NVIDIA AI stack, on-prem GPU infrastructure, or open-source LLM serving.
• Contributions to open-source AI/agent projects, relevant publications, or cloud/AI architecture certifications.

In this role, success means a secure, reusable AI Factory — built once and adopted across domains — where agents ship faster on shared, well-governed platform services. Architecture decisions are trusted by engineers and business partners alike, solutions run reliably and cost-effectively in production, and Advantech is able to adopt frontier AI safely and ahead of the curve.