ML Platform Lead Engineer

What matters most

• AWS-Centric ML Platform Infrastructure via IaC

  The primary objective is architecting, implementing, and securing cloud-native Data Science and AI/ML infrastructure on AWS. This directly involves managing core services like SageMaker, Bedrock, and EKS using reusable Terraform modules and automation scripts in Python.

• Automated End-to-End CI/CD and Container Orchestration

  A core duty is establishing scalable CI/CD pipelines using GitLab (or tools like Jenkins, GitHub Actions, Azure DevOps) to automate testing and release workflows. This pairs with managing containerized environments using Docker and Kubernetes (Amazon EKS) with Helm charts and rolling deployments.

• Federated Operating Model and Standards Adoption

  Because LSEG uses a federated model with core teams in London, Bucharest, and Bangalore alongside engineers embedded in business divisions, a key responsibility is driving standard DevOps culture, platform engineering best practices, and operational excellence across distributed groups.

• Enterprise Cloud Migration, Reliability, and Governance

  Supporting the migration of existing applications to AWS while maintaining strict governance is essential. The role demands implementing security best practices, comprehensive logging, monitoring, and disaster recovery strategies tailored for enterprise financial services.

Likely challenges to think through

• Balancing Strict Financial Governance with Data Science Agility

  Enterprise financial infrastructure mandates rigorous security, IAM controls, and compliance, which can create friction against data scientists needing rapid experimentation with SageMaker and Bedrock. Prepare to discuss how to design self-service guardrails, secure VPC patterns, and automated IAM policies that ensure compliance without stalling innovation.

• Enforcing Consistency Across a Federated Architecture

  With core engineering dispersed across three geographic hubs and engineers embedded in business divisions, standardizing CI/CD pipelines and Terraform modules is difficult due to varying local team habits. Prepare strategies for building modular, versioned platform tooling and driving adoption through developer advocacy and shared governance.

• Managing High-Throughput Kubernetes and ML Workload Lifecycle

  Deploying ML services and models on Amazon EKS requires balancing autoscaling, resource efficiency (such as GPU/compute provisioning), and zero-downtime rolling deployments. Prepare to demonstrate deep technical mastery in Helm chart design, cluster ingress, networking configurations, and cluster reliability under variable computational loads.

Topics to prepare

• Cloud-Native AI/ML Infrastructure Architecture on AWS

  Relevance: The posting explicitly tasks the role with designing and securing AWS cloud infrastructure for Data Science and AI/ML across SageMaker, Bedrock, EC2, VPC, IAM, S3, RDS, and EKS. Exercise: Diagram an enterprise ML platform topology showing VPC isolation, private subnets, IAM role least-privilege configurations, secure S3 data storage, and connectivity between SageMaker studio/endpoints and foundational models in Bedrock.

• Container Orchestration and Kubernetes Reliability on Amazon EKS

  Relevance: Requires hands-on experience running Kubernetes (Amazon EKS) with Helm charts, ingress controllers, networking, autoscaling, and rolling deployments. Exercise: Review architectural concepts of Horizontal Pod Autoscaling (HPA) based on custom inference metrics, ingress controller setup with ALB/NLB, and blue/green or rolling deployment strategies to avoid service interruption during model and platform updates.

• Modular Infrastructure as Code (IaC) and State Governance

  Relevance: Explicitly requires 3+ years of provisioning AWS infrastructure via Terraform, developing reusable modules, and handling remote state. Exercise: Review the conceptual lifecycle of Terraform state locking using S3 and DynamoDB, dependency management between core shared platform networks and tenant environments, and module versioning patterns across multiple environments.

• Enterprise CI/CD and MLOps Release Engineering

  Relevance: The role requires building and optimizing automated build, test, deployment, and release pipelines using GitLab CI/CD, Jenkins, Azure DevOps, and GitHub Actions. Exercise: Map out an end-to-end automated deployment pipeline integrating linting, unit testing, container vulnerability scanning, infrastructure plan generation, and gated multi-stage deployment to staging and production clusters.

• Platform Engineering and Federated Operating Models

  Relevance: LSEG outlines a federated model where core platform teams in London, Bucharest, and Bangalore empower embedded engineers across divisions to drive adoption and standardization. Exercise: Formulate a strategy for balancing centralized governance (security controls, base Helm templates, shared Terraform modules) with business division autonomy, ensuring smooth developer enablement without bottlenecking innovation.

Skills to revise

• Writing Reusable AWS Terraform Modules with Remote State

  Relevance: The job posting mandates 3+ years of Terraform experience creating reusable modules and managing remote state for AWS infrastructure. Exercise: Write a clean Terraform module that provisions a private EKS nodegroup or VPC configuration, including strict input variables, output attributes, and an S3/DynamoDB remote state backend configuration.

• Authoring Multi-Stage GitLab CI/CD Automation Pipelines

  Relevance: Required qualifications highlight creating scalable CI/CD automation solutions using GitLab CI/CD or equivalent tools. Exercise: Write a .gitlab-ci.yml file that automates a container build: lint Python code, run automated tests, build and push a Docker image to Amazon ECR, and trigger an automated rolling Helm upgrade.

• Creating Helm Charts and Configuring Autoscaling for Kubernetes Deployments

  Relevance: Demands strong experience with Docker, Helm charts, Kubernetes networking, ingress controllers, autoscaling, and rolling updates. Exercise: Package a containerized microservice into a Helm chart with parameterized values.yaml defining an Ingress resource, readiness/liveness probes, and an HPA spec with autoscaling thresholds.

• Developing Python Automation and Operational Tooling

  Relevance: Explicit requirement for 5+ years of developing automation scripts, deployment utilities, and infrastructure tooling using Python. Exercise: Write a Python utility script utilizing boto3 to audit AWS IAM roles or S3 bucket policies across an ML platform, or automate a operational health-check and alerting hook for model endpoints.

• Configuring Model Serving and Security Integrations with AWS SageMaker and Bedrock

  Relevance: Strong hands-on experience with AWS AI/ML services—specifically SageMaker and Bedrock—is explicitly listed as a primary required skill. Exercise: Use boto3 or Terraform to script the deployment and invocation of an Amazon SageMaker real-time endpoint or call an Amazon Bedrock foundation model with VPC endpoint constraints and IAM role verification.

• Prepare an end-to-end ML platform architecture walkthrough on AWS

  Completion means delivering a concise 5-minute technical narrative and architectural diagram illustrating an enterprise Data Science/ML platform. It should clearly highlight VPC network isolation, private endpoints for Amazon SageMaker and Amazon Bedrock, IAM role configurations, Amazon S3 data lakes, and Amazon EKS cluster components.

• Build a modular Terraform repository demonstrating remote state and AWS provisioning

  Completion looks like presenting a functional, structured Terraform project featuring reusable modules (e.g., EKS nodegroups or secure VPC networks), parameter validations, and an S3 backend with DynamoDB state locking to demonstrate 3+ years of IaC best practices.

• Package a sample containerized service using Docker and Helm on Kubernetes

  Completion means having a working Dockerfile and Helm chart repository that includes parameterized deployment manifests, Ingress definitions, readiness/liveness probes, and Horizontal Pod Autoscaler (HPA) settings to validate hands-on Kubernetes orchestration experience.

• Construct a multi-stage CI/CD deployment pipeline configuration

  Completion looks like authoring a production-ready GitLab CI (or GitHub Actions/Azure DevOps) YAML configuration demonstrating automated testing, container image build and vulnerability scanning, and gated Terraform deployment stages.

• Assemble evidence of Python infrastructure automation and tooling

  Completion involves preparing 1-2 code samples or architectural examples showcasing automation utilities built with Python and boto3, such as automated cluster cleanups, infrastructure health monitors, or migration tooling.

• Formulate STAR-format examples for cross-functional collaboration in a federated model

  Completion means documenting 2-3 structured stories (Situation, Action, Result) illustrating how you partnered with distributed engineering and architecture teams to drive DevOps adoption, infrastructure standardization, or legacy cloud migrations.

• Draft thoughtful questions about LSEG's AI/ML Transformation Program

  Completion means preparing 3-4 specific questions focused on the operating reality, such as how platform governance is split between core engineering (London/Bucharest/Bangalore) and embedded divisional teams, and the current migration roadmap for foundational models in Bedrock.