Senior AI Engineer (Applied)

What matters most

• Own AI evaluation and benchmarking

  Design frameworks for text, vision and generative AI systems; set metrics and thresholds, then analyze failures across quality, latency, reliability and cost.

• Deliver reusable production AI services

  Architect scalable, observable services, evaluation tools and microservices for multiple product teams, from system design through deployment and iteration.

• Build reliable generative AI systems

  Design RAG pipelines and agentic workflows that account for performance, scalability and operational reliability.

• Make experiments reproducible and decisions evidence-based

  Establish experimentation, versioning and service-reliability practices. Work with product managers to define success criteria, assess feasibility and evaluate feature trade-offs.

• Lead integration and technical direction

  Integrate AI systems with software and data engineering teams; assess emerging tools and recommend build, partner or hybrid approaches using evidence. Mentor junior engineers and explain decisions to varied stakeholders.

Likely challenges to think through

• Balancing model quality with service constraints

  Likely challenge, not a confirmed employer problem: an approach that benchmarks well may miss latency, reliability or cost thresholds in production. Prepare to explain how you would set thresholds, inspect failure cases and decide when a quality gain is worth the operational cost.

• Making shared tools useful across product teams

  Likely challenge, not a confirmed employer problem: reusable services need consistent interfaces and observability without blocking different gaming and platform use cases. Prepare an architecture example showing what you would standardize, what you would leave configurable and how you would test integration.

• Choosing emerging AI approaches without chasing novelty

  Likely challenge, not a confirmed employer problem: new GenAI and agentic tools may offer capabilities without clear evidence of production value. Prepare a comparison method using product success criteria, reproducible benchmarks and operational trade-offs to justify a build, partner or hybrid recommendation.

Topics to prepare

• Evaluation and statistical reasoning

  Review how to choose metrics, set thresholds and interpret experimental results. As a practice exercise, sketch a benchmark for a hypothetical text, vision or generative AI feature, including how you would investigate failures in quality, latency, reliability and cost. The posting explicitly assigns ownership of these evaluations.

• Production AI service trade-offs

  Compare design choices for a scalable, observable service used by multiple product teams. As a practice exercise, draw an architecture and explain what you would monitor and how latency, reliability and cost would affect the design. The role calls for production-grade AI services and reusable platforms.

• RAG and agentic workflow design

  Review the stages and failure points of RAG pipelines and agentic workflows. As a practice exercise, diagram a hypothetical workflow and identify where you would evaluate output quality, performance and operational reliability. Both system types are named in the responsibilities.

• Experimentation and reproducibility

  Review how versioning and repeatable experiments support dependable AI decisions. As a practice exercise, specify what you would record so another engineer could reproduce a benchmark and compare a later iteration. The posting calls for practices in experimentation, versioning and reproducibility.

• Product and sourcing decisions

  Review how success criteria and technical evidence inform feasibility and build-versus-partner-versus-hybrid decisions. As a practice exercise, prepare a short decision rationale with explicit trade-offs for a hypothetical AI feature. The role involves product-manager collaboration and data-backed recommendations.

Skills to revise

• Implement an evaluation harness

  Practice writing Python code that runs a small set of test cases, calculates selected quality and service metrics, applies thresholds and groups failures for review. This directly reflects the posting’s Python, benchmarking and failure-analysis responsibilities.

• Build and test an AI backend API

  Practice defining an AI-service endpoint, separating its components into modules and writing tests for expected responses and errors; FastAPI is one example named in the posting. This revises the required API design, backend-service and testing fundamentals.

• Prototype a RAG or agentic flow

  As a preparation exercise, build a small RAG pipeline or agentic workflow and test how it handles unsuccessful outputs and slow steps. The posting explicitly calls for building these systems under performance and reliability constraints; the prototype is not a stated employer project.

• Package and deploy an AI service

  Practice packaging a service with Docker and outlining a deployment on one of the named cloud platforms, including where Kubernetes orchestration would fit. Rehearse how you would check service health after deployment. The posting asks for cloud deployment experience and Docker/Kubernetes familiarity.

• Explain decisions across teams

  Practice presenting one benchmark result twice: a technical explanation for software and data engineers, then a concise success-criteria and trade-off summary for a product manager. This reflects the role’s cross-functional communication and senior-level decision-making responsibilities.

• Prepare an end-to-end AI project walkthrough

  The role owns work from experimentation through deployment and iteration. Prepare one real NLP, vision, or generative AI example showing your decisions, collaborators, results, and what you changed after release; label any hypothetical elements.

• Draft an evaluation and benchmarking plan

  For a sample text, vision, or generative AI feature, write proposed quality, latency, reliability, and cost metrics; thresholds; test cases; and a failure-analysis method. Explain how you would compare results and account for uncertainty.

• Sketch a production AI service

  Draw a proposed architecture for a gaming or platform use case, including APIs, observability, scaling, and integration with other product systems. If using RAG or agents, show the workflow and its reliability and performance trade-offs.

• Gather implementation and deployment evidence

  Identify real examples, if available, of Python code, modular API design, tests, AI backend services, cloud deployment, and Docker or Kubernetes use. Be ready to distinguish hands-on work from familiarity.

• Practice a product trade-off recommendation

  Prepare a brief hypothetical decision memo with proposed success criteria, feasibility risks, and a data-backed way to compare building an AI capability, partnering, or using a hybrid approach.

• Rehearse leadership and communication examples

  Prepare specific accounts, if available, of collaborating with product, software, or data teams; mentoring junior colleagues; and explaining a technical decision to a non-technical stakeholder.

• Write targeted questions for Razer

  Ask how this team selects gaming and platform use cases, sets evaluation thresholds, shares AI services across product teams, and measures reliability after deployment. Clarify expectations for work in the Singapore office and the stated occasional travel.