Research And Development Engineer

Research Engineer — On-Premises AI Appliance & Agentic Design Wizard

Full-stack engineering · AI/agent systems · UI design

NUS Department of Computer Science · Singapore · Full-time

About the role

We build on-premises, air-gapped AI appliances ("Studios") for regulated industries — financial institutions and semiconductor manufacturers — where data cannot leave the building. The product is a sealed box that runs agentic AI locally, plus an Agentic Design Wizard that takes a customer's specification and automatically generates and customises the appliance's interface for their domain.

We're looking for a Research Engineer — a full-stack builder who can also design — to own both the appliance's web application and the wizard that produces it. You'll work at the intersection of research and product: turning agentic-AI research into shipped, usable systems, end-to-end from interface design through front-end build, back-end services, and integration with the on-device AI runtime. The UI/UX sensibility to make complex AI workflows genuinely usable matters as much as the engineering.

This is a high-ownership role on a small team. You'll ship the core product, not a feature on someone else's.

What you'll build

• The appliance application — own the full-stack web app that runs on the Box: a React/Next.js front end and a Python/FastAPI back end, built to run entirely on-device.
• The Agentic Design Wizard — design and build the system that takes a specification as input and automatically generates and customises the appliance's UI/UX, guiding configuration through a structured, multi-stage setup flow. This is the centrepiece of the product and the most novel part of the role.
• AI integration — wire the front end and back end to the on-device agentic runtime: LLM orchestration (e.g. LangGraph), multi-agent workflows, retrieval-augmented generation (vector database + embeddings), and locally-served language models. Handle streaming, grounding, and graceful behaviour when models are uncertain.
• UI/UX design and implementation — own the appliance's interface design: information architecture, interaction design, and a reusable design system in Figma — then build it to production quality. Translate dense AI/agent workflows and outputs into clear, usable views.
• On-premises delivery — package and deploy for air-gapped, security-sensitive environments: containerisation, reproducible builds, and engineering that respects the constraints of regulated customers.
• Iteration — work with the AI/research team to turn agent capabilities into product features, and refine the experience based on customer and usability feedback.

Requirements

Required

• Strong full-stack experience: React + Next.js + TypeScript on the front end, Python + FastAPI (or equivalent) on the back end.
• Demonstrated UI design ability — you can design interfaces (Figma, design systems), not just implement others' designs. Portfolio or work samples required.
• Hands-on experience building AI-powered product features — integrating LLMs/AI APIs, retrieval (RAG), prompt engineering, and streaming responses into real applications.
• Solid software-engineering fundamentals: API design, state management, Git, testing, and deployment.
• A proven ability to take ambiguous or complex requirements and turn them into working, polished products end-to-end.

Preferred

• Experience with agentic frameworks (LangGraph, LangChain) and multi-agent orchestration.
• Vector databases, embeddings, and RAG pipelines (e.g. Qdrant, pgvector, BGE/sentence-transformer models).
• Self-hosted / local LLM serving (vLLM, Ollama, or similar) and running models on NVIDIA GPU hardware (CUDA).
• Containerisation and on-premises deployment (Docker), ideally in air-gapped or security-sensitive settings.
• Spec-to-UI / code-generation / low-code or design-system-driven UI generation experience — directly relevant to the wizard.
• Data visualisation skills (D3.js or charting libraries) for analytical, data-rich views.
• Domain exposure to financial services, regulated industries, or industrial/semiconductor manufacturing.

Bonus

• Experience building guided "wizard" or configuration flows.
• Self-hosted observability / evaluation tooling (e.g. LangSmith) for LLM applications.
• Some HCI/UX research exposure (usability testing, user interviews).

Why join

• Own the core of a genuinely novel product — an AI appliance and the wizard that builds its interfaces.
• Work across the full stack and the full lifecycle: design, build, ship, iterate.
• Small, technical team where your decisions shape the product directly.
• Frontier work at the intersection of agentic AI, on-premises systems, and product design.

How to apply

• Send your CV, portfolio/work samples, and (if available) a link to relevant code to [email protected], with the subject line "Research Engineer — AI Appliance".