Data Science Manager (CXD)
Job Description:
Rakuten Group, Inc. is a global leader in internet services and has a diverse ecosystem spanning across e-commerce, fintech, communications and more serving approximately 1.8 billion members worldwide. Founded in Tokyo in 1997, the Group operates in over 30 countries and regions with more than 30,000 employees.
Based in Singapore's Central Business District, Rakuten Asia Pte. Ltd. serves as the regional headquarters for Asia, driving value through areas such as advertising product development, product strategy, and data management to support Rakuten Group's global ecosystem. Learn more at:
The Data Science Manager will lead the applied AI science function within Rakuten's CXD team, managing a team of data scientists and ML engineers responsible for the intelligence layer of our conversational AI platform. Based in Singapore and working closely with product and engineering counterparts across Japan and the wider APAC region, you will set the technical direction, build team capability, and drive measurable improvement across model quality, retrieval performance, and personalisation — ensuring that science directly moves business outcomes at scale.
This role is chartered around two foundational responsibilities that define the scientific integrity of the platform as it scales across multiple Business Units Runtime Decision Plane (Mathematical & Scientific Sign-off) and Learning Plane (Flywheel & Feedback Systems).
Key Responsibilities
Team Leadership & Science Strategy: Build, manage, and mentor a team of data scientists and ML engineers. Define the team's technical roadmap across both the Runtime Decision Plane and the Learning Plane, establish ways of working, and create an environment where rigorous science and rapid iteration coexist. Set the scientific bar for the group-wide Customer Service AI platform — enabling rapid, safe onboarding of distinct Business Units beyond the initial deployments. Partner closely with Product and Engineering leads across Singapore and Japan to align the science agenda with platform priorities and BU expansion commitments.
LLM Strategy & Model Lifecycle Management: Own the end-to-end lifecycle of large language models in production — including vendor evaluation, model selection, fine-tuning, and version upgrade planning. Shift the team's focus from basic prompt optimisation to systematic model comparison frameworks and agentic reasoning evaluation, grounding every production decision in formal theory and empirical benchmarking rather than heuristic tuning. Provide scientific sign-off on orchestration strategies, multi-step agent routing policies, and chain-of-thought reasoning design — establishing a clear, defensible process for every model change that reaches production. Govern the cost-quality-latency trade-off across all BU deployments.
RAG Architecture & Continuous Iteration: Lead the ongoing research and improvement of Retrieval-Augmented Generation systems, covering retrieval quality, reranking strategies, knowledge base health, and generation faithfulness. Drive a structured iteration cycle — hypothesis, offline evaluation, staged rollout, impact measurement — and maintain a clear improvement backlog prioritised against business impact. Distinguish retrieval gaps from LLM synthesis failures to ensure the evaluation framework correctly attributes root causes. Collaborate with Knowledge Engineering teams to ensure retrieval and content quality reinforce each other across all BU knowledge bases.
Evaluation Framework, Experimentation & VoC Intelligence: Define and operate a multi-layer evaluation framework spanning offline benchmarks, online A/B experiments, and continuous production monitoring. Define and enforce pre-launch BU evaluation gates — statistical coverage thresholds, intent accuracy floors, hallucination rate ceilings — that every new BU deployment must pass before going live. Own the team's Voice of Customer analysis practice — building error taxonomies and active learning pipelines that transform production telemetry and VoC signals into structured learning cycles. Hold the bar for measurement rigour across all product changes, ensuring the Learning Plane operates as a systematic feedback flywheel rather than a series of one-off eval runs.
Conversational Memory, Context & User Profile Management: Set the technical direction for the memory and context layer — session state, long-term user profile modelling, and contextualised retrieval. Ensure the system delivers coherent, personalised interactions at scale while meeting privacy and compliance requirements. Work with engineering to translate architectural decisions into scalable production systems.
AI-Driven Personalisation for Upsell & Conversion: Partner with Product and Business stakeholders to define and deliver personalisation capabilities that shift the chatbot's role from a support tool to an active growth channel — covering user segmentation, intent prediction, eligibility-aware recommendation, and guided selling signal design. Build the scientific framework for customer-centric personalisation grounded in behavioural signals and lifetime value thinking, not just answer quality. Own the measurement framework that connects personalisation investments to conversion and revenue outcomes, and coach the team to reason about the problem from customer operations and LTV lens.
Mandatory Requirement
7+ years in applied machine learning or NLP, including 3+ years in a people management or technical lead role
PhD in Computer Science, Machine Learning, Natural Language Processing, or a related technical discipline strongly preferred; exceptional candidates with equivalent research-level industry experience will be considered
Deep hands-on experience with LLMs and RAG systems in production — retrieval pipeline design, reranking, chunking strategy, knowledge base management
Demonstrated ability to provide mathematical and algorithmic sign-off on system design decisions — formally justifying and defending orchestration strategies, routing policies, and retrieval scoring mechanisms, not only implementing them
Experience with agentic reasoning evaluation, multi-step orchestration analysis, and systematic model comparison frameworks — moving beyond prompt tuning to platform-level scientific rigour
Strong track record designing and running online experiments (A/B, interleaving, bandits) and translating results into product decisions
Experience building error taxonomies, evaluation frameworks, and feedback loops that systematically improve production quality metrics — not just one-off eval runs
Experience building and developing high-performing data science teams; comfortable with hiring, performance management, and career development
Demonstrated ability to synthesise analytical findings into clear product problem statements and influence prioritisation at a leadership level
Proficiency in Python and cloud ML infrastructure (GCP preferred); familiarity with vector search and embedding models
Strong communication skills — able to represent the science function credibly to both technical and non-technical senior stakeholders across cultures and time zones
Comfortable working in a cross-regional environment with key stakeholders based in Japan
Rakuten is an equal opportunities employer and welcomes applications regardless of sex, marital status, ethnic origin, sexual orientation, religious belief, or age.