Staff Generative AI Engineer

Building custom model architectures, optimising training loops, and deeply understanding library internals to enhance performance for production GenAI systems. You'll be mentoring others on best practices.

Designing and building complex, production-grade agentic systems and RAG pipelines. You'll understand their failure modes and build custom components to overcome limitations, setting standards for their use.

Designing and optimising indexing strategies for performance and cost in large-scale RAG systems. You'll manage data ingestion pipelines and evaluate trade-offs between different indexing algorithms.

Architecting and automating end-to-end MLOps pipelines on these platforms. This includes managing fine-tuning jobs, endpoint scaling, and cost optimisation for large-scale GenAI deployments.

Designing standardised experiment tracking templates for teams, building custom dashboards, and synthesising findings from hundreds of experiments to inform architectural decisions.

Writing efficient, multi-stage Dockerfiles and deploying/managing GenAI services on Kubernetes. You'll understand concepts like pods, services, and deployments to ensure scalable and resilient infrastructure.

Managing complex codebases, leading branching strategies, resolving intricate merge conflicts, and conducting thorough code reviews for team members. You'll be setting our Git best practices.

• Product Managers (for defining requirements and features)
• Engineering Leads (for system integration and MLOps)
• Data Scientists (for model experimentation and evaluation)
• Security & Compliance Teams (for ensuring safe and ethical AI deployment)
• Senior Leadership (for strategic updates and resource allocation)

• Cloud Providers (e.g., AWS, Google, Azure for platform capabilities)
• Open-Source Communities (for contributing and drawing insights)
• Research Partners (occasionally, for collaborative projects)

The quality of your GenAI outputs is directly tied to the quality of your data. As systems become more complex, the need for sophisticated data pipelines—especially for unstructured text, images, and audio—is paramount for training, fine-tuning, and RAG.

Large-Scale Data Curation & Cleaning · Active Learning & Data Labelling Strategies · Data Governance & Lineage for AI · Real-time Data Streaming for RAG

• This month: Audit our current data pipelines for a key GenAI product, identifying bottlenecks and areas for quality improvement.
• Next quarter: Design and implement an improved data curation pipeline for a specific GenAI use case, focusing on automation and quality checks.
• Within 6 months: Lead the integration of a real-time data streaming solution to enhance the freshness of one of our RAG systems.
• Throughout: Explore tools like Apache Spark, Flink, and data observability platforms tailored for AI data.

Quick win: Improve the chunking strategy and metadata extraction for one of our existing RAG data sources. Small changes here can have a big impact.

Privacy concerns and regulatory pressures mean we can't always centralise all data. Federated learning and other privacy-preserving techniques will become crucial for training and fine-tuning models on sensitive or distributed datasets without compromising privacy.

Federated Averaging Algorithms · Differential Privacy · Homomorphic Encryption · Secure Multi-Party Computation (MPC)

• This month: Read introductory papers on Federated Learning and Differential Privacy to grasp the core concepts.
• Next quarter: Prototype a small federated learning experiment using a framework like TensorFlow Federated or PySyft.
• Within 6 months: Evaluate the feasibility of applying privacy-preserving AI techniques to one of our internal GenAI use cases involving sensitive data.
• Throughout: Monitor industry developments and regulatory guidance on privacy-preserving AI.

Quick win: Identify a small, internal dataset that could benefit from privacy-preserving techniques and explore how differential privacy could be applied during data pre-processing.