AI platform engineer

An AI platform engineer makes AI something people across our client’s organization can use on their own. The job is removing gatekeeping without giving up control. Developers get self-service access to build AI-powered features, and the rest of the organization gets the same for the AI tools they use day to day. Either way, the platform still enforces who can reach what, what it costs, and what data is allowed to leave the building. That is the infrastructure and governance underneath enterprise AI coding assistants, cloud-native AI platforms, and enterprise AI subscriptions used across the organization, not a single model or a research project.

Key Responsibilities

Infrastructure and Scaling

• Provision AI compute: stand up and maintain the compute environment AI workloads run on, including containerized and GPU-backed resources, so workloads run efficiently and cost-effectively

• Automate safe rollouts: build automated pipelines for rolling out AI platform updates, with rollback built in

Model Access and Orchestration

• Build retrieval systems: connect AI models to the organization’s own data so responses are grounded in real, current information

• Maintain vector stores: connect and maintain the vector stores and data sources retrieval depends on

• Route and gate model access: give people one governed way to reach approved AI models and providers, routing each request to the model that actually fits the job

Security and Governance

• Redact sensitive data: strip sensitive or private data out of prompts before they reach a model

• Build guardrails: put guardrails in place to catch harmful, biased, or simply wrong AI output before a user sees it

• Enable self-service without losing control: set up access and permissions so people across the organization can use approved AI capabilities on their own, while the platform still enforces where the line is

• Monitor system health: watch system health, response times, and error rates, and catch problems before people have to report them

• Platform engineering experience: 4+ years of hands-on experience in platform engineering, cloud infrastructure, or DevOps

• AI in production: experience running AI or LLM platforms in production, including retrieval systems that ground model answers in real company data

• AI systems judgment: understands that AI systems behave differently from regular software, and knows how to test, evaluate, and monitor them so problems get caught early

• Programming and automation: strong Python skills, including building integrations against APIs

• Cloud and containers: solid experience with at least one major cloud provider (AWS or Azure) and building on containerized infrastructure

• Cost and model judgment: some sense of cost and performance tradeoffs between AI models, and when a smaller, cheaper model is the right call

• Reporting and transparency: experience building or exposing metrics through APIs or dashboards for reporting and visibility beyond the immediate team

• Security mindset: builds with security as a first-class concern: least-privilege access, careful handling of sensitive data, sensible defaults

• Self-service thinking: understands how to make AI capability self-service, and where the guardrails actually need to hold

• Full-stack comfort: equally comfortable in infrastructure, automation, integration, and application code

Preferred Qualifications

• Terraform or similar infrastructure-as-code experience

• Exposure to cloud cost management or FinOps practices

• Deeper experience with retrieval pipelines: chunking, embeddings, indexing, and ranking, not just calling a vector database

• Experience routing requests across multiple AI model providers based on cost, speed, or capability

• Judgment about when a simple, predictable workflow beats an open-ended AI agent, with experience building both

• Familiarity with container orchestration for AI or compute-heavy workloads

• Familiarity with governing tool access for AI agents, such as scoped permissions or a central tool registry, as agentic workflows become more common