Head, MLOps - Singapore - TECHNOLOGY SERVICES GROUP PTE. LTD.

Description

Job Description & Requirements

Our company is at the forefront of innovation in financial technology. Our mission is to develop an AI-driven platform that predicts market trends, empowering retail investors with the tools previously only accessible to professionals. We're looking for a talented MLOPs Software Engineer with an interest in finance to join our growing team in our Singapore office.

Responsibilities

• Focuses on MLOps from model training, model evaluation to trading evaluation.
• Develop large-scale deployment of GPU nodes running in dozens of Kubernetes clusters across regions. Core technologies may include: CUDA, Python, PyTorch, Triton, Redis, NCCL, NVLink.
• Develop simulation pipelines to scale the testing of hyperparameters for model training and for trading. The innovation here lies in ability to synchronize model parameters with financial trading parameters.
• Automate feature evaluation and experimentation for model training.
• Automate model training and model performance evaluation.
• Combine classical model evaluation with financial trading evaluation metrics such as Sharpe, Turnover, Fitness, Returns, Drawdown, Margin, Long Count, Short Count, Sector/Subindustry Allocations etc.
• Build, maintain, enhance model pipelines using MLOps and AIOps frameworks.
• Work with full suite of data and model infrastructure and integration i.e. SageMaker Studio or Vertex A.I.
• Work with Data Scientists to setup model monitoring and feedback systems.
• Part of a sub-team focusing on full autonomous AI Trading under a larger trading division.

Requirements:

• Bachelor's and above. Software Engineering background would be preferred.
• Must have stack: CUDA, PyTorch, low level languages such as C++ or Go.
• Must have experiences :
• High scale distributed training and inference systems and ML systems.
• Parallel and distributed computing with multiple gpus at once
• High throughput scheduling as a service particularly at supercomputingscale .
• Understand how to resolve Segmentation Faults
• Must have testing, debugging experiences: Unit Tests, Integration Tests, Errors Handling to ensure trading system remains fault tolerance and robust
• Must have orchestration tools: DAGs in Airflow or Prefect, MLFlow
• Must have deployment experiences: Docker, AWS, Load Testing etc
• Good to have network technologies inside and out