About the Beijing AI Center

The Beijing AI Center is a new strategic investment by AstraZeneca to accelerate drug discovery through AI. The center brings together AI researchers, computational scientists, and platform engineers to apply foundation models, agentic AI, and large-scale scientific computing to real R&D problems. Situated in one of the world's most dynamic AI talent markets, the center operates at the intersection of AI and biologics discovery, computational chemistry, and data-driven drug development.

The center is structured around three pillars: Discovery verticals (biologics engineering, computational chemistry) that own the science, Data & AI Platforms (this role) that own the capabilities, and R&D IT that owns the infrastructure. A dedicated on-premises GPU cluster provides the compute backbone, operated by IT and shaped by platform standards.

About the Role

This role owns the AI capability layer for the Beijing AI Center: the methods, standards, data products, evaluation frameworks, and AI application platforms that sit between raw compute infrastructure and Discovery science teams. You define how AI gets done in Beijing. Discovery teams decide what science to pursue. IT operates the cluster. You build the reusable platform capabilities that connect the two.

You are not an infrastructure leader. IT handles GPU provisioning, cluster operations, networking, and hardware vendor relationships. You are not a science leader. Discovery teams own model architecture decisions, training objectives, and scientific interpretation. You own the methodology, tooling, data readiness, and evaluation rigor that make both sides more productive.

What You Will Do

AI Methodology & Standards (20%)

Methods and best practices (not model architecture, which Discovery owns)

• Define and maintain AI engineering standards for the center: training configurations, experiment reproducibility requirements, naming conventions, checkpoint management
• Curate and document reusable training recipes: distributed training templates, parameter-efficient fine-tuning configurations, and inference optimization patterns
• Publish best-practice guidance on method selection (when to use parameter-efficient methods vs. full fine-tuning, when to apply alignment techniques) so Discovery teams can make informed choices
• Establish reproducibility framework: experiment configs, seed management, result validation criteria

Boundary with Discovery: Discovery scientists select model architectures, define training objectives, curate domain-specific training data, and interpret results. You provide the methods toolkit they draw from and the engineering standards they follow. You build the car; they drive it.

Benchmarking & Evaluation (15%)

• Design and build the center's model evaluation platform: standardized test harnesses, metric dashboards, and comparison tooling
• Curate domain-relevant benchmark suites in partnership with Discovery teams (biologics, computational chemistry)
• Define evaluation methodology: what metrics matter, how to construct held-out test sets, how to avoid data leakage, when a model is "good enough" to deploy
• Build leaderboard infrastructure so Discovery teams can rigorously compare model variants, fine-tuning approaches, and architectural choices
• Establish evaluation-as-a-service: Discovery teams submit a model, get back a standardized scorecard

Boundary with Discovery: Discovery teams define what success looks like for their scientific problems and interpret evaluation results. You build the evaluation infrastructure and methodology that makes rigorous comparison possible.

Data Products & Governance (20%)

• Own the "AI-ready" data layer: transform IT's foundational data pipelines into governed, versioned, documented datasets purpose-built for model training
• Build data product catalogue with clear provenance, quality scores, known limitations, and recommended use cases
• Implement data quality frameworks: automated validation rules, drift detection, completeness checks, cross-dataset consistency
• Design and enforce data versioning standards so experiments are reproducible against specific dataset snapshots
• Implement PIPL/DSL compliance framework for cross-border data transfer and local data operations
• Define data governance policies aligned to Enterprise Data Enablement FAIR standards

Boundary with IT: IT builds and operates the data transfer pipelines, storage, and base ETL. You define what "AI-ready" means and build the quality, versioning, and governance layer on top.

Agentic AI & Applications (15%)

• Deploy agentic AI capabilities for the Beijing center: multi-agent orchestration, local agent development frameworks, scientific workflow automation
• Evaluate and integrate China-specific LLM landscape (Qwen, DeepSeek, GLM, Moonshot) for center use cases
• Build China cloud integrations (Alibaba, Tencent, Baidu) for LLM inference and agent hosting
• Enable AI productivity tooling for Discovery teams: coding assistants, literature search agents, experiment design copilots
• Define agent evaluation methodology: how to measure agent reliability, safety, and scientific accuracy

Boundary with IT: IT provides the hosting infrastructure for agent services (containers, endpoints, networking). You own what gets built, which LLMs are selected, and how agents are evaluated.

Team Leadership (20%)

• Recruit, manage, and develop a team of 8 platform FTEs in 2026, scaling to 10+
• Set technical direction and quality bar for all platform deliverables
• Run the hiring pipeline in Beijing: source candidates from top local universities and the Beijing AI talent market
• Create a team culture that attracts top Beijing AI talent in a competitive market
• Onboard and develop junior hires; establish technical mentorship with global platform leads

Stakeholder Alignment (10%)

• Partner with IT infrastructure lead on compute policies, scheduling priorities, and capacity planning (you provide requirements and priorities; IT implements)
• Work directly with Discovery vertical leads to understand workload requirements and prioritize platform delivery
• Represent platform capabilities to ecosystem partners (academic institutions, AI companies) where technical integration is needed
• Provide regular delivery updates to global AI leadership
• Participate in weekly coordination meetings across all center functions

Requirements

Experience

• 10+ years in AI/ML platform engineering, data engineering, or applied AI at scale
• 5+ years leading technical teams (8+ direct reports minimum)
• Track record building platform capabilities from scratch (not just maintaining)
• Experience building internal AI products or platforms that research or science teams actually adopted
• Experience serving scientific or research teams with platform capabilities (biopharma, genomics, or similar domain preferred)

Technical

• Strong background in AI training methods: distributed training, parameter-efficient fine-tuning, and modern post-training and alignment techniques
• Experience designing evaluation frameworks and benchmarking infrastructure for ML models
• Data engineering and governance: data product design, quality frameworks, versioning, compliance
• Modern AI application development: LLM tooling, agent frameworks, prompt engineering, multi-agent orchestration
• MLOps fundamentals: experiment tracking, model registry, CI/CD for ML

China-Specific

• Ability to work in Beijing full-time (on-site)
• Mandarin fluency required (team, stakeholders, and partners operate in Mandarin)
• China cloud provider experience (Alibaba Cloud, Tencent Cloud, Huawei Cloud)
• Familiarity with Chinese AI ecosystem: local LLMs (Qwen, DeepSeek, GLM), academic institutions, AI companies
• PIPL/DSL awareness (will learn on the job if foundational understanding exists)

Leadership

• Proven ability to operate in a matrix organization: you report to global AI leadership but deliver locally for the center
• Comfortable with ambiguity: the center is still being built, and priorities will evolve
• Strong stakeholder management across Discovery scientists, IT infrastructure, and global platform teams
• Product mindset: you measure success by adoption and impact, not just delivery
• Talent magnet: ability to attract top AI talent in a competitive Beijing market

Nice-to-Have

• Biopharma domain knowledge (drug discovery, protein engineering, computational chemistry)
• Experience with model evaluation at scale (leaderboards, automated evaluation pipelines)
• Academic partnership management (joint projects, co-supervision, IP frameworks)
• Experience navigating multi-org delivery models where platform, infrastructure, and science are separate teams

Date Posted

10-4月-2026

Closing Date

AstraZeneca embraces diversity and equality of opportunity. We are committed to building an inclusive and diverse team representing all backgrounds, with as wide a range of perspectives as possible, and harnessing industry-leading skills. We believe that the more inclusive we are, the better our work will be. We welcome and consider applications to join our team from all qualified candidates, regardless of their characteristics. We comply with all applicable laws and regulations on non-discrimination in employment (and recruitment), as well as work authorization and employment eligibility verification requirements.