Principal Software Engineer (AI / LLM) - 首席軟體工程師

What you’ll do:

- Own design & development of agent workflows that span email, SMS, social, and contact-management microservices.
- Collaborate daily with product managers, data scientists, and fellow engineers to turn marketing use-cases into resilient, observable code.
- Mentor junior engineers on evaluation-driven development and modern agent patterns.
- Stay current on fast-moving LLM infrastructure (vLLM, LangGraph, MCP) and guide the team in adopting what works.
- Shape long-term architecture, ensuring the platform can plug into future AI ecosystems without lock-in.


You will be a good fit if you:

- Enjoy freedom and take responsibility in a remote first environment. (You could choose WFH or in Taipei office)
- Are excellent at communicating clearly. (In English of course)
- Know how to effectively and efficiently collaborate with people around the world (cross timezones).
- Can manage multiple projects concurrently and prioritize effectively.
- Leverage open source technologies in a smart and practical way to solve problems while creating maintainable and evolving systems.
- Enjoy maintaining a high standard and discussing code quality and design patterns with peers.
- Flexible attitude and willingness to frequently move between different teams, software architectures and priorities.
- Desire to collaborate with our other product teams to think strategically about how to solve problems.

Must-Have Qualifications:

- Backend Engineering – 8 + years designing and operating REST / gRPC microservices; expert Python & FastAPI; strong SQL / NoSQL data modelling.
- Cloud & DevOps – Hands-on production AWS (Lambda, EKS / Fargate, S3) and infrastructure-as-code with Terraform, CloudFormation, or Helm.
- Messaging & Streaming – Practical experience with SQS queues and Kinesis streams for event-driven systems.
- AI & Data – Working knowledge of LLMs, prompt engineering, retrieval-augmented generation (RAG), embeddings, and vector databases such as Pinecone, Milvus, or Weaviate.
- Kubernetes – Able to author Helm charts, configure autoscaling / HPA, and monitor clusters with Prometheus & Grafana.


Preferred Qualifications:

- Built or maintained high-throughput, low-latency data systems.
- Familiar with MLOps fundamentals (feature stores, model registries) and data-privacy best practices.
- Created automated LLM evaluation harnesses using OpenAI Evals, LangChain Bench, MLflow LLM Evaluation, or DeepEval.
- Designed red-team and guardrail tests for safety, bias, PII leakage, and prompt-injection resilience with Promptfoo / Guardrails.
- Hands-on with LangGraph and other Python agent frameworks (LangChain Agents, CrewAI, AutoGen); designed multi-step agent workflows, branching logic, and reliability guards.
- Implemented Model Context Protocol (MCP) servers and adapters to let AI agents securely access external data and tools; defined context schemas, capability negotiation, RBAC, audit logging, and integrated MCP tools via LangGraph / LangChain MCP adapters.
- Mentored engineers and contributed to architecture reviews, design documents, and code reviews.

Updated: 2025/06/28

Video and on-site interview in English