AI Full-Stack Engineer (Next.js / AI Agent)

Vlight Technology Limited; Hong Kong; On-site; HKD 480,000 - 640,000; Full time

Job summary

Vlight Technology Limited in Hong Kong is seeking a skilled full‑stack developer to lead AI agent and generation features using Mastra, Vercel AI SDK, Next.js, and Tailwind.

You will design robust data models in PostgreSQL, implement tool calling, and ensure high-quality, cost-aware AI deployments while writing clear technical docs.

Qualifications

• Strong full-stack fundamentals with React/Next.js and TypeScript.
• Hands-on LLM app experience with prompts, tool calling, or streaming.
• Solid PostgreSQL schema design and SQL; experience with Supabase.
• Ability to review AI-generated code and ensure final quality.
• Fluent English reading and clear technical writing.

Responsibilities

• AI Agent development: build AI agents and tool calling.
• AI generation: create image, video and text generation features.
• AI quality and cost: model evaluation, tracing, and provider selection.
• Full-stack development: ship frontend and backend with Next.js, Supabase, shadcn/ui and TailwindCSS.
• Engineering productivity: use AI coding tools to boost speed and quality.

Skills

TypeScript

React

Next.js

LLM experience

PostgreSQL

SQL

Supabase

AI tooling

Code review

English

Documentation

Tools

Vercel AI SDK

Mastra

LangChain

OpenAI SDK

Trigger.dev

ffmpeg

Job description

Responsibilities

• 1. AI Agent development: build AI agents and tool calling with Mastra and the Vercel AI SDK.
• 2. AI generation: develop image, video and text generation features, orchestrating background jobs with Trigger.dev.
• 3. AI quality and cost: own model evaluation, tracing, and model selection across providers (OpenAI / Anthropic / Google).
• 4. Full‑stack development: ship frontend and backend features with Next.js, Supabase, shadcn/ui and TailwindCSS.
• 5. Engineering productivity: use AI coding tools such as Claude Code and Codex to keep raising development speed and quality.

Requirements

• 1. Strong full‑stack fundamentals: expert in TypeScript, React and Next.js (App Router / Server Components / Server Actions); able to take a feature from design to production on your own.
• 2. Hands‑on LLM application experience: shipped a real project with at least one of Vercel AI SDK / Mastra / LangChain / OpenAI SDK; understand prompt design, tool calling, streaming and structured output.
• 3. Solid PostgreSQL schema design and SQL; experience with Supabase or a similar BaaS.
• 4. Heavy user of AI coding tools who reviews and verifies AI‑generated code and owns the final quality.
• 5. Fluent English reading; good communicator with strong ownership and self‑drive; writes clear technical docs.

Nice to have

• 1. Experience with agent memory, RAG / vector search (pgvector, embeddings, rerank) or multi‑agent systems.
• 2. Built LLM evaluation (evals, LLM‑as‑judge) or production AI observability (traces, cost and latency analysis).
• 3. Worked on image / video generation products, or media processing with tools like ffmpeg.
• 4. Deployed on Vercel, Cloudflare Workers or Google Cloud; used Trigger.dev or a similar background job framework.
• 5. Built SaaS products (Clerk auth, Stripe subscription billing, i18n), or know hardware, manufacturing or supply chains.