Developer (AI Agent)

Pulsar; Hong Kong; On-site; HKD 300,000 - 520,000; Full time

Job summary

Pulsar is hiring a Developer to design, build and operate long-running agent services in production. You will craft agent infrastructure, CI/CD and deployment pipelines, and integrate tool calls within frontier models.

You will also shape safety controls, memory systems and evaluation tooling to ensure reliability, cost efficiency and scalability. The role emphasizes working with retrieval systems, RAG and knowledge graphs, while maintaining a strong focus on security and maintainability in a

Qualifications

• 3-5 years of software engineering experience, incl. building LLM-based applications or agents.
• Degree in Computer Science, Software Engineering or related discipline.
• Proficiency in Python, TypeScript, or Rust on Linux.
• Experience shipping agentic systems using tool calling and multi-step loops.
• Understanding of LLM security risks and how to design permission boundaries.
• Experience with retrieval systems (vector databases, embeddings, RAG).
• Familiar with evaluation and tracing tooling for LLM systems (Langfuse, LangSmith, OpenTelemetry).
• Focus on reliability, scalability, cost efficiency and maintainability of deliverables.
• Strong interest in financial markets and blockchain technologies.
• Experience with MCP, agent SDKs, or high-throughput system design is a plus.

Responsibilities

• Design, build and operate long-running agent services in production, including agent infrastructure, CI/CD and deployment.
• Build agent loops and tool-calling integrations on frontier models such as Claude and GPT.
• Design agent safety and permission controls, including prompt injection protection and access control at account and tool levels.
• Build agent memory and knowledge systems using RAG, knowledge graphs and embeddings.
• Own the agent evaluation pipeline, including tracing, observability and regression testing of agent behaviour.
• Engineer the agent harness: context management, long-session reliability and cost control.

Skills

LLM-based apps

3-5 years experience

Python/TypeScript/Rust

Linux environments

agent loops

LLM security

retrieval systems

evaluation tooling

Education

Computer Science or Software Engineering degree

Tools

Langfuse

LangSmith

OpenTelemetry

Vector databases

Job description

• You are pioneering and innovative and want to be part of the cutting-edge and disruptive crypto-currency world


• You are excited about bringing AI agents out of demos and into real production systems


• You are eager to learn new knowledge in both financial and technical fields


• You thrive in a non-hierarchical organization with a casual working environment


• You are flexible and dedicated. We operate an on-call rota around the clock to ensure our global operations run smoothly

Team / Role

As a Developer, you will:

• Design, build and operate long-running agent services in production, including agent infrastructure, CI/CD and deployment


• Build agent loops and tool-calling integrations on frontier models such as Claude and GPT


• Design agent safety and permission controls, including prompt injection protection and account- and tool-level access control


• Build agent memory and knowledge systems using RAG, knowledge graphs and embeddings


• Own the agent evaluation pipeline, including tracing, observability (e.g. Langfuse) and regression testing of agent behaviour


• Engineer the agent harness: context management, long-session reliability and cost control

Required skillset

You should get in touch if:

• You have 3-5 years of software engineering experience, including hands-on work building LLM-based applications or agents


• You hold a university degree in Computer Science, Software Engineering or a related discipline


• You write clean, production-quality code in Python, TypeScript, Rust or similar, and are familiar with Linux environments


• You have shipped agentic systems using tool calling and multi-step agent loops, not only single-prompt applications


• You understand LLM security risks such as prompt injection and data exfiltration, and how to design permission boundaries against them


• You have experience with retrieval systems (vector databases, embeddings, RAG; graph-based retrieval is a plus)


• You have built or used evaluation and tracing tooling for LLM systems (e.g. Langfuse, LangSmith, OpenTelemetry)


• You own the reliability, scalability, cost efficiency and maintainability of what you deliver


• You show a strong interest in financial markets and blockchain technologies


• Experience with MCP, agent SDKs, or high-throughput system design is a plus

We encourage applicants to read through our Privacy Notice for Applicants before submitting your applications - https://pulsar.com/careers/