The Role

As a Machine Learning Engineer, you will be responsible for deploying and retraining models and ensuring that they are functional in production. You will not be required to build custom models, but you will need to have strong knowledge of machine learning concepts in order to successfully deploy deep learning models into production. 

The data science team will build the models, but they may not be production-ready, hence you will need to optimize the models to ensure the algorithms can be run in the cloud. Additionally, you will need to know how to tune machine learning systems in the cloud due to the relatively low latency limitations. 

You will be working with data engineers, data scientists, backend software engineers and DevOps engineers to ensure that the deployed models are scalable and performant.

Reporting to: The Machine Learning Lead     

Required Skills

• Experience with ML frameworks, PyTorch is preferred.

• Experience with cloud computing (AWS Lambda, AWS Sagemaker, AWS S3)

• Knowledge of dockerising deep learning models (e.g. Docker, TorchServe, AWS MMS).

• The use of automated ML solutions (e.g. GCP AutoML / Azure AutoML) DOES NOT constitute as experience.

Preferred Skills 

• Experience with runtime inference optimization (e.g. TensorRT, AWS Inferentia, Quantization, ONNX, GPU-based inference).

• Experience with deploying dockerized deep learning models to the cloud (e.g. Sagemaker endpoints, Vertex AI docker images). 

• Experience with code as IaC is a plus. (eg. AWS CDK, Terraform)

What you will be doing

• Work with the engineering team to develop and tune deep learning models.

• Build monitoring solutions to ensure that deep learning instances are optimized.

• Problem-solving and project management.

• Procure MLOps practices to manage model versioning and region-specific inferences.

• Undertake any other duties relevant to the position as instructed by your immediate supervisor.

• You will also be required to write some python scripts to help orchestrate the data flow between different hosted ML endpoints.