OCBC Indonesia | Data Science Team | Jakarta, Indonesia
Build the future of intelligent banking. On our own GPU clusters.
You're not just a data scientist. You're an AI enabled data scientist who builds LLM-powered intelligence that actually works in the real world, especially in high-stakes financial environments. At OCBC Indonesia, you'll design and deploy production-grade RAG systems that help operations, risk analysts, and customer service teams make faster, smarter, and more compliant decisions. And unlike most companies that rely on cloud APIs, you'll work directly on our in-house GPU clusters, giving you rare, hands-on access to low-level AI infrastructure, distributed inference, and model optimization at scale.
This is not a research lab. This is a high-impact AI team inside one of Southeast Asia's most trusted banks, where your work directly reduces risk, accelerates loan approvals, and improves customer experience. All with full ownership and visibility.
Key Responsibilities
- Design, build, and deploy production-grade RAG systems to enhance banking operations, customer support, and risk documentation. Integrating structured financial data with unstructured text (loan applications, contracts, regulatory docs, customer chats).
- Own end-to-end LLM-powered workflows: from query understanding and retrieval optimization to response generation, grounding, and post-processing. Ensuring outputs are accurate, compliant, and auditable.
- Develop and maintain robust evaluation frameworks (evals) to measure RAG performance across dimensions: relevance, faithfulness, answer correctness, latency, and hallucination rate, using both automated metrics (e.g., RAGAS, LLM-as-a-Judge) and human-in-the-loop validation.
- Collaborate with Data Engineering and MLOps to integrate RAG pipelines into production systems, leveraging vLLM for high-throughput, low-latency inference at scale.
- Apply prompt optimization frameworks like GEPA (Generative Evaluation and Prompt Architecture) to systematically improve LLM responses. Iterating on templates, chain-of-thought structures, and retrieval strategies based on eval feedback.
- Partner with Product, Risk, and Compliance to translate business needs into AI requirements.
- Drive A/B testing and offline/online evaluation of LLM-enhanced features, measuring uplift in approval rates, reduction in manual review time, or improvement in customer NPS.
Minimum Qualifications
- Bachelor's or master's degree in computer science, statistics, mathematics, engineering, or a related quantitative field.
- 4+ years as a Data Scientist or ML Engineer in banking, fintech, or financial services, with at least 1 year focused on LLMs and RAG systems.
- Strong proficiency in Python and core ML libraries, with experience in LLM orchestration frameworks.
- Proven hands-on experience designing, implementing, and evaluating RAG pipelines. including vector databases, embedding models, retrieval ranking, and response grounding techniques.
- Expertise in LLM evaluation frameworks. Building custom eval suites, defining metrics (e.g., precision@k, faithfulness, context relevance), and using tools like LangChain, LlamaIndex.
- Experience with end-to-end ML lifecycle: data wrangling, feature engineering, model training, offline/online evaluation, A/B testing, and production deployment.
- Ability to communicate complex AI concepts clearly to business, risk, and engineering teams, with a bias for clarity over jargon.
Preferred Qualifications (A Plus!)
- Hands-on experience with vLLM for high-throughput LLM inference in production environments.
- Practical application of GEPA (Generative Evaluation and Prompt Architecture) or similar prompt optimization frameworks to systematically improve LLM performance.
- Experience with LLM fine-tuning (LoRA, QLoRA), instruction tuning, or distillation for domain-specific financial use cases.
- Familiarity with financial data domains: credit risk modeling, loan underwriting, KYC/AML documents, regulatory text, or financial NLP.
