New opportunity

AI Software Engineer

Razer · Singapore

About this role

Razer seeks an AI Software Engineer to design, develop and deploy scalable AI-powered solutions for its gaming ecosystem. The role integrates AI and machine learning models into applications and optimizes production workflows in collaboration with data scientists, software engineers, DevOps teams and cloud engineers.

Responsibilities

Implement and containerize production AI/ML models, optimizing speed, accuracy and cost. Deploy models through RESTful APIs, gRPC or edge computing solutions.
Design secure, performant and documented APIs and microservices. Build AI pipelines for real-time and batch processing, and improve AI API and plugin architectures.
Monitor production model performance, logging and errors; maintain API latency and availability against service-level agreements. Apply MLOps practices, including CI/CD for machine learning workflows on local machines and cloud platforms.
Turn research models into deployable solutions, coordinate deployments across teams and evaluate new AI/ML frameworks and API or plugin development approaches.

Requirements

3+ years of AI/ML software development experience and a proven record of deploying AI solutions to production. Strong model deployment and API development experience using RESTful APIs, gRPC or edge computing solutions.
Strong Python, Java or C++ programming; hands-on experience with TensorFlow, PyTorch and Scikit-learn; and in-depth knowledge of Docker, Kubernetes and DevOps practices.
Experience deploying AI applications at the edge with TensorRT, ONNX and TFLite and on AWS, Azure or Google Cloud Platform, using Terraform and cloud AI services such as AWS SageMaker, Google Vertex AI and Azure ML.
Experience with SQL, NoSQL and vector databases; knowledge of distributed computing with Spark or Dask and parallel processing. Strong written and verbal communication, adaptability and willingness to learn in a fast-paced agile environment.
A bachelor’s or master’s degree in computer science, AI or a similar discipline from an accredited institution.

The role is based in Razer’s Singapore office and may require up to one travel trip per year.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

AI model deploymentMachine learningMLOpsPythonJavaC++TensorFlowPyTorchScikit-learnDockerKubernetesRESTful APIsgRPCMicroservicesCI/CDEdge computingTensorRTONNXTFLiteAWSAzureGoogle Cloud PlatformTerraformAWS SageMakerGoogle Vertex AIAzure MLSQLNoSQLVector databasesSpark (Apache)

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Production model deployment
Implement and containerize AI/ML models with Docker and Kubernetes, optimizing speed, accuracy and cost. The posting calls for a track record of deploying AI solutions to production.
Reliable AI services
Design secure, performant, documented APIs and microservices using RESTful APIs or gRPC; monitor model performance, errors, latency and availability against required SLAs.
Scalable pipelines and MLOps
Build real-time and batch AI pipelines and implement CI/CD for machine-learning workflows on local machines and cloud platforms.
Edge and cloud delivery
Develop and deploy AI applications at the edge and on cloud infrastructure; the posting names TensorRT, ONNX, TFLite, Terraform and major cloud AI services.
Cross-team integration and improvement
Work with data scientists to turn research models into deployable solutions, coordinate with DevOps and cloud engineers, and identify improvements to AI API and plugin architectures.

Likely challenges to think through

Balancing serving constraints
Likely challenge, not a confirmed problem: improving model accuracy may increase latency or cost while APIs must meet availability SLAs. Prepare to explain how you would benchmark alternatives, set operating targets and monitor the deployed service.
Choosing between edge and cloud
Likely challenge, not a confirmed problem: edge deployments can constrain compute and model size, while cloud delivery introduces infrastructure and service costs. Prepare a decision framework covering latency, resource limits, deployment and observability.
Moving research into production
Likely challenge, not a confirmed problem: a research model may need reproducible packaging, integration and error handling before release. Prepare a concrete approach to coordinating handoffs, CI/CD and rollback with data science, software and DevOps teams.

Practice your answers

Technical practice

How would you turn a research model into a production AI service?

What to coverSuggested preparation: Explain how you would validate model behavior, package its dependencies and define an inference interface. Discuss how you would work with data scientists to preserve accuracy while making a TensorFlow, PyTorch or Scikit-learn model deployable.

How would you balance inference speed, accuracy and cost?

What to coverSuggested preparation: Describe the measurements you would gather before changing a production model or its serving configuration. Explain how you would evaluate improvements against accuracy, latency and operating cost rather than optimizing one measure alone.

How would you containerize and scale a model deployment?

What to coverSuggested preparation: Discuss how Docker packaging and Kubernetes deployment affect reproducibility, resource use and scaling. Explain how you would investigate a deployment that works locally but performs poorly when containerized.

How would you design an AI API or plugin integration?

What to coverSuggested preparation: Compare RESTful APIs and gRPC for an appropriate inference interface, including security, performance and documentation. Discuss how you would assess changes to an existing API or plugin architecture without disrupting its consumers.

When would you deploy inference at the edge rather than in the cloud?

What to coverSuggested preparation: Compare latency, device constraints, connectivity and operational cost. Explain how you would test model compatibility and accuracy when considering TensorRT, ONNX or TFLite for edge deployment.

How would you deploy an AI application on cloud infrastructure?

What to coverSuggested preparation: Explain what you would provision with Terraform and how you would assess a managed AI service such as SageMaker, Vertex AI or Azure ML. Address the trade-offs between managed deployment and greater control over the serving infrastructure.

How would you design real-time and batch AI pipelines?

What to coverSuggested preparation: Contrast throughput, latency and failure handling for the two processing modes. Explain when parallel or distributed processing with Spark or Dask would be useful, and how you would check that it improves the pipeline.

What would you include in CI/CD for a machine-learning workflow?

What to coverSuggested preparation: Outline checks for code, model artifacts and deployment changes across local and cloud environments. Explain how you would make releases reproducible and detect a model or API regression before it reaches production.

How would you diagnose an AI API missing its latency or availability SLA?

What to coverSuggested preparation: Walk through the monitoring, logs and error signals you would examine to separate model-serving issues from API or infrastructure issues. Describe how you would restore service and verify that model performance and availability remain acceptable.

How would you choose storage for an AI application's data?

What to coverSuggested preparation: Compare SQL, NoSQL and vector databases by the data and query patterns they support. Explain how you would evaluate retrieval performance, data consistency and operational complexity for a proposed use case.

Behavioural practice

Tell me about a time you worked with a data scientist to deploy a research model.

What to coverSuggested preparation: Use a real situation, your actions and the result to show how you resolved differences between research assumptions and production needs. Make your own contribution and the deployment outcome clear.

Describe a deployment you coordinated across engineering teams.

What to coverSuggested preparation: Choose a real situation involving coordination with software, DevOps or cloud engineers. Explain your actions to align responsibilities and communicate risks, then state the result.

Tell me about a production performance or availability problem you handled under pressure.

What to coverSuggested preparation: Set out the situation, the evidence you used and the actions you took to diagnose and address it. Describe the result, including what you learned about monitoring or error handling.

Describe a time you learned a new technology and applied it to your work.

What to coverSuggested preparation: Give a real example that explains why the technology was needed in a fast-moving project. Cover how you evaluated it, what you did and the result, without overstating its impact.

Tell me about a time you proposed an improvement to an existing API or integration.

What to coverSuggested preparation: Describe the situation, how you identified the opportunity and how you communicated its trade-offs to others. Explain what action was taken and the result, including any limits to what was achieved.

Sharpen your knowledge

Topics to prepare

Production model trade-offs
The posting calls for optimizing deployed models for speed, accuracy and cost. Revision suggestion: compare two deployment configurations and explain how you would measure and choose between those three outcomes.
API architecture and service levels
The role covers secure, performant, documented APIs and microservices, including RESTful APIs or gRPC, with latency and availability SLAs. Revision suggestion: sketch a model-serving API and explain its interface, error handling and response to an SLA breach.
MLOps and production observability
The posting specifies ML CI/CD, monitoring, logging and production model performance. Revision suggestion: map a research model’s path to deployment, then identify checks before release and signals to monitor afterward.
Edge and cloud deployment choices
The role names edge deployment with TensorRT, ONNX and TFLite, and cloud infrastructure and AI services on AWS, Azure or Google Cloud Platform. Revision suggestion: compare an edge and a cloud deployment for the same model, noting operational and performance trade-offs.
Real-time, batch and distributed processing
The posting calls for real-time and batch AI pipelines, database management, and knowledge of Spark or Dask and parallel processing. Revision suggestion: diagram both processing paths and identify where SQL, NoSQL or vector data storage and distributed work might fit.

Skills to revise

Prepare a model for serving
The posting asks for strong Python, Java or C++ skills and hands-on use of TensorFlow, PyTorch and Scikit-learn. Practice suggestion: use a small model in a named framework to write an inference path, test its outputs and measure inference time; this rehearses the move from model to deployable solution.
Build and document an inference API
The role explicitly includes RESTful APIs or gRPC and documented, secure microservices. Practice suggestion: expose a small model through one of those interfaces, document inputs and outputs, and test invalid requests and error responses.
Containerize and release a service
Docker, Kubernetes, DevOps and ML CI/CD are named in the posting. Practice suggestion: containerize the inference service, specify how it would run in Kubernetes, and outline CI/CD checks and a deployment rollback; these address production deployment.
Instrument model serving
The role calls for monitoring, logging, error handling, and meeting API latency and availability SLAs. Practice suggestion: add request-error and latency logging to the service, then inspect a simulated failure and explain what you would monitor for degraded model performance.
Rehearse edge and cloud deployment
The posting names TensorRT, ONNX, TFLite, Terraform and cloud AI services. Practice suggestion: test a model’s edge export with one named format or runtime, then draft a Terraform-based cloud deployment plan using one named provider and its AI service. This is a revision exercise, not a stated employer project.
Work through a data pipeline
The role specifies batch and real-time pipelines, SQL, NoSQL and vector databases, plus Spark or Dask knowledge. Practice suggestion: process a small batch with Spark or Dask, describe its parallel steps, and write sample reads or writes for the relevant named database types.

Your preparation checklist

Prepare a production deployment walkthrough
The posting asks for a proven record of deploying AI solutions. If you have a relevant example, prepare a diagram and brief account of your contribution, model framework, deployment path, and measured speed, accuracy and cost trade-offs; do not invent results.
Sketch a model-serving design
As interview preparation, draw a secure, documented microservice that serves a model through REST or gRPC. Explain containerization, error handling, scaling and how you would test latency and availability against an SLA; the posting does not prescribe a specific design exercise.
Map out an MLOps release and monitoring plan
Prepare a one-page flow from research model to production using CI/CD, Docker and Kubernetes where appropriate. Show how you would monitor model performance, logs and failures, and coordinate a deployment with data scientists and DevOps engineers.
Compare edge and cloud deployment choices
The role covers both edge and cloud AI applications. Prepare a short comparison of when you would use an edge format or runtime such as ONNX or TensorRT versus a cloud deployment, including latency, resource and cost considerations; distinguish hands-on evidence from proposed approaches.
Work through a pipeline and data scenario
For practice, outline real-time and batch processing paths, appropriate uses of SQL, NoSQL or vector storage, and when parallel processing with Spark or Dask might help. Label this as a hypothetical design unless it reflects work you actually did.
Select evidence for the stated prerequisites
Make a concise evidence sheet covering relevant production deployments, programming in Python, Java or C++, ML frameworks, infrastructure as code and cross-team communication. Note which examples demonstrate direct experience and which areas you would discuss conceptually.
Draft questions about the role’s priorities
Prepare questions about the AI APIs or plugins needing improvement, current latency and availability targets, how edge and cloud work is divided, and how teams move research models into production. These are questions to ask, not assumptions about Razer’s current systems.

Availability last checked: 4 Nov 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in Singapore. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…