AI Research Engineer

221b Limited (“221b”) develops an anti-crime platform that uses AI and network analysis to support investigations into financial fraud, anti-money laundering (AML), and communications activity.

Job Summary:

We are hiring an AI Research Engineer for a permanent, full-time R&D role in Hong Kong on a confidential Agentic AI project for financial services. Agentic AI here means LLM-based software that plans multi-step tasks and uses approved tools within defined limits. You will run experiments to improve evidence retrieval, answer quality, explainability, multilingual support, reliability, speed, and cost, then turn successful results into improved systems and prototypes. You will report to the R&D lead and work with the engineering team.

Responsibilities:

• Plan and run experiments for AI systems that analyse financial-crime data and complete multi-step tasks.

• Evaluate suitable hosted LLMs and system designs using controlled tests and measurable criteria.

• Build and test multi-step AI workflows with approved tool use, validation rules, and human approval where needed.

• Research and test Retrieval-Augmented Generation (RAG), which supplies relevant source documents to the model. Work will cover document preparation, chunking, embeddings, reranking, search, and vector databases.

• Create benchmarks for answer accuracy, source support, unsupported claims, difficult inputs, explainability, and multilingual/CJK performance. Use component comparisons and error analysis to identify what works.

• Test system architectures to improve response time, request capacity, reliability, scalability, and cost.

• Build proofs of concept, record results and limitations, and work with engineers to move validated research into the platform.

Requirements:

• Relevant Bachelor’s, Master’s, or Doctoral degree in a recognized STEM discipline such as Computer Science, Data Science, Artificial Intelligence, or Machine Learning. A Master’s degree is preferred; strong candidates with a relevant Bachelor’s degree will also be considered.

• Experience planning ML / AI experiments, comparing results quantitatively, and writing clear technical reports.

• Strong proficiency in Python and experience building APIs or AI-enabled software services.

• Hands-on experience with RAG, multi-step AI systems that call tools, prompt design, or AI system evaluation.

• Good understanding of machine learning, natural language processing, statistics, and experiment design.

• Experience with LLM APIs, vector databases, LangGraph, LangChain, LlamaIndex, or similar tools.

• Ability to explain research methods, results, limitations, and conclusions clearly in English.

Nice-to-Haves:

• Experience with prompt design, structured outputs, tool calling, or systematic LLM evaluation.

• Experience testing multi-step AI systems, tool permissions, validation checks, and human approval steps.

• Experience with IBM watsonx.ai, OpenShift, Kubernetes, or repeatable deployment and testing environments.

• Experience with multilingual NLP, particularly Chinese, Japanese, or Korean datasets.

• Previous R&D work in financial services, financial-crime prevention, or another regulated industry.

What We Offer:

• Flexible working hours and work-from-home policy.

• Access to professional AI development tools.

• On-the-job training and technical guidance from senior engineers.

• Opportunity to develop and test Agentic AI and LLM systems using enterprise deployment and monitoring tools.

Job Highlights:

• Flexible working hours and work-from-home policy

• Full-time R&D on Agentic AI and LLM technology for financial-services investigations

• Hands-on work in multi-step AI workflows, RAG, evaluation, and system integration