AI Engineer (Agents, Context & Evals) [VT26-AIE-ACE] [Hong Kong] [Full Time]
Company:
Votee AI
Location:
Hong Kong
Job Type:
Full-Time
About Votee AI
At Votee AI, we are pushing the boundaries of artificial intelligence. We build and deploy state-of-the-art AI systems to solve some of the most complex problems for highly regulated, mission-critical organizations. We are looking for an elite Forward Deployed Engineer to lead the on-site implementation of our systems for top-tier clients across the public sector, statutory regulatory bodies, and critical public infrastructure.
The Role
We're hiring an
AI Engineer
to own the reliability layer of the LLM systems we ship — the agent harness, context pipelines, and evaluation infrastructure that turn raw model calls into software a regulator can read.
We hire for skill, not pedigree. If you've built real things with agents — professionally, on your own time, or both — we want to talk.
Responsibilities
- Design and maintain the context layer for our agents, including retrieval pipelines, prompt assembly, tool schemas, memory, and session state.
- Practice spec-driven development (OpenSpec-style): write executable specifications that define agent behavior, then build both the agent and its eval suite from those specs.
- Build and maintain eval suites as first-class code — prompt-level tests, regression suites, LLM-as-judge pipelines, and production traces — and gate releases on their results.
- Design and operate the agent harness, covering tool boundaries, retry and fallback policy, OpenTelemetry tracing, token and latency budgets, and circuit-breakers for runaway tool-use loops.
- Lead harness-side response on Sev-1 incidents: diagnose the root cause, write the postmortem, and land the harness or eval change that prevents recurrence.
- Build for regulated deployment from day one, with auditability primitives, output provenance, and human-in-the-loop checkpoints baked into the harness rather than bolted on.
- Raise the team's floor on AI-native engineering by contributing shared tooling, reusable patterns, and harness primitives that every future engagement inherits.
Requirements
- A non-trivial project you've shipped with an LLM or agent stack — professional, personal, or hackathon work all count.
- Hands-on with at least one modern agent stack: Claude Agent SDK, OpenAI Agents SDK, Google ADK, LangGraph/LangChain, or an in-house harness.
- Context engineering beyond prompt-tweaking — retrieval, tool schemas, memory, session state.
- Can articulate how you'd prove an agent's output is trustworthy.
OpenCode
,
Cursor
, or equivalent AI-native developer tooling.
- Strong Python and software fundamentals. Working English required; Cantonese is a plus, not a requirement.
Nice-to-Haves
- Experience with MCP (Model Context Protocol), A2A (Agent2Agent Protocol).
- Pre-LLM ML or evaluation-methodology background.
- Exposure to regulated or compliance-sensitive environments.
- Open-source contributions to agent, tracing, or eval tooling.
Why Join Votee AI?
You will be trusted with massive responsibility from day one. You will get to work with the latest AI technologies, directly impact critical infrastructure, and be part of an agile, fast-moving startup environment where your engineering skills will be pushed to the bleeding edge.
To Apply To Apply: Please submit your resume along with your GitHub/Portfolio. If you have examples of systems you've built using LLMs or multi-agent frameworks, please include them!