AI Engineer (1 Year Contract)

What matters most

• Run and improve the global AI sandbox

  The role helps develop and maintain the sandbox, integrate AI solutions, support teams experimenting in it, and improve it using user feedback and technological advances.

• Develop and refine AI solutions

  The posting calls for developing and deploying AI models, analysing data for actionable insights, and fine-tuning models for accuracy, performance and scalability.

• Build quality and risk alignment into experiments

  AI products and experiments must align with Prudential’s risk, data protection and technology risk management policies and applicable regulations. The engineer also contributes to quality assurance.

• Guide product incubation with current AI research

  The engineer conducts R&D on AI, machine learning and LLM frameworks, then provides technical guidance to teams developing products through the incubation program.

• Work across teams and develop AI capability

  The role collaborates with technology, AI and business stakeholders and external partners, mentors junior engineers, and contributes to education and knowledge sharing.

Likely challenges to think through

• Balancing experimentation with safeguards

  Preparation hypothesis: making a shared sandbox useful for rapid experiments while keeping those experiments aligned with risk and data protection requirements may require careful boundaries. Prepare to explain how you would define permitted data use, review points and evidence of compliance without assuming Prudential’s specific controls.

• Choosing between model quality and operational performance

  Preparation hypothesis: fine-tuning for accuracy can compete with scalability or performance. Prepare an example of how you would set evaluation criteria, compare alternatives and justify a deployment decision using measured trade-offs.

• Turning research into usable product guidance

  Preparation hypothesis: emerging frameworks may not suit every incubating team’s needs. Prepare to describe how you would assess a new approach against a product goal, communicate limitations to technical and business stakeholders, and decide whether further experimentation is warranted.

Topics to prepare

• Data preparation and model evaluation

  Revise how preprocessing, feature engineering and evaluation choices affect model results. Exercise: outline a validation plan for a hypothetical dataset, identifying leakage risks and metrics you would use to judge accuracy. The posting explicitly calls for knowledge of these areas.

• LLM, NLP and computer vision frameworks

  Compare when an LLM, an NLP model or a computer vision model would be appropriate, and explain how you would assess each. Exercise: make a short comparison of relevant framework capabilities and limitations. The posting names these framework areas but does not prescribe a particular use case.

• Model optimization trade-offs

  Revise the relationship between accuracy, performance and scalability. Exercise: explain how you would diagnose a model that meets an accuracy target but runs too slowly, then rank possible improvements and their evaluation criteria. Model optimization and fine-tuning are stated responsibilities.

• Sandbox, cloud and model operations

  Review how an experimental model moves into a maintainable sandbox environment, including the distinct concerns of MLOps and LLMOps. Exercise: sketch an integration and monitoring plan and explain what you would check before changing a shared experiment. The role supports a global AI sandbox and calls for cloud-platform familiarity.

• Risk, data protection and quality assurance

  Revise how to identify data-handling, technology-risk and model-quality issues during AI experimentation. Exercise: draft a review checklist with a test or evidence item for each issue. The posting requires alignment with Prudential’s policies and applicable regulations, without specifying individual controls.

• AI product incubation and technical guidance

  Consider how to turn experimental findings into recommendations for technology, AI and business teams. Exercise: outline how you would explain a model trade-off and next steps to an incubating product team. Technical guidance, cross-functional collaboration and mentoring junior engineers are stated duties.

Skills to revise

• Prepare and analyse data in Python and R

  Practice cleaning a small sample dataset, creating features and summarising findings in Python and R. Be ready to explain why each transformation was made. The posting names both languages and asks for actionable insights from data.

• Build a reproducible model comparison

  Using scikit-learn, PyTorch or TensorFlow, train a small baseline and an alternative, record the preprocessing and evaluation choices, and compare results. These libraries are named examples, and developing and deploying AI models is an explicit requirement.

• Run a focused framework experiment

  As a preparation exercise, try a small NLP, computer vision or LLM-framework experiment and document its inputs, outputs, failure cases and suitability for further incubation. The posting cites all three framework areas and expects AI research and experimentation; it does not mandate this particular exercise.

• Integrate an experiment into a sandbox design

  Draft the steps to make a model available in a shared sandbox, including how a team would use it and how feedback would inform changes. Discuss where AWS, Google Cloud or Azure familiarity would help, without assuming a specified service or architecture. Sandbox maintenance and integration are core duties.

• Test a model change before deployment

  Take a model with a hypothetical accuracy or runtime problem, propose a tuning change, and prepare a before-and-after test table covering accuracy, performance, scalability and quality checks. The role explicitly calls for optimization, fine-tuning and quality assurance.

• Communicate a technical recommendation

  Write a brief recommendation for a business and engineering audience that states the evidence, limitations, data-protection questions and proposed next experiment; then practise explaining it as guidance to a junior engineer. This reflects the posting’s collaboration, technical-support and mentoring responsibilities.

• Map evidence to the stated requirements

  Prepare a one-page reference to relevant degree, five-plus years of AI model development and deployment, Python and R work, framework experience, and cloud familiarity. These are requirements in the posting; use only evidence you can substantiate.

• Build a model-to-deployment walkthrough

  Select one real project and document the problem, data preprocessing, feature engineering, model choice, evaluation, deployment, and outcome. Be ready to explain your own decisions and trade-offs.

• Sketch a global AI sandbox design

  As a practice exercise, draw how teams could experiment with AI applications and integrate solutions into a shared sandbox. Explain maintenance, user support, and how feedback would inform improvements; the posting does not specify an architecture.

• Prepare an optimization case study

  Show a real example of measuring and improving accuracy, performance, or scalability. Include baseline results, experiments, trade-offs, and any applicable MLOps or LLMOps practices.

• Draft a risk and quality review

  For a hypothetical AI experiment, outline questions to check data protection, technology risk, applicable regulations, and quality assurance. The posting calls for alignment with Prudential’s policies but does not provide their details.

• Rehearse technical guidance examples

  Prepare concise examples of explaining an AI recommendation to business and technology stakeholders, supporting a product team, and mentoring a junior engineer. Note where external-partner collaboration was involved, if applicable.

• Write targeted questions for the employer

  Ask which sandbox improvements are priorities, how incubating teams receive technical support, how model quality and risk alignment are reviewed, and what outcomes matter most during the one-year contract.