VP - AI Engineering - name

Job Description:

The VP - AI Engineering owns the central AI engineering platform — the “AI Harness” — together with the agent frameworks, evaluation infrastructure, and AI-assisted development practices that every CSI engineering team builds on. A small central team builds the platforms, connectors, and plumbing across LLMs, data, knowledge, and workflows; functional teams build on top of them. This role is accountable for a hard delivery-velocity number, for the engineering bar on AI- and agent-generated code, and for the developer enablement that makes adoption real rather than nominal.

Own the delivery-velocity mandate

Set and hit the ~50% delivery-acceleration target: establish the Year-0 baseline and the DORA-style metrics (lead time for change, deployment frequency, change-failure rate, time to restore), and report progress to the CDAO and pillar peers.

Translate individual adoption into organizational throughput — redesign team scope and workflow around human judgment plus agent execution, not individual task completion.

Build the AI Harness as a product

Own the strategy and roadmap for the central AI platform — model routing, connectors, MCP servers, and the build-vs-buy framework that decides when to use Bedrock or Foundry rather than direct LLM access.

Treat the Harness as a product, not infrastructure — named services, SLAs, adoption metrics, and a backlog driven by engineering-team needs — while managing token and cost optimization across LLM usage.

Empower developers and agents

Build and maintain the agent framework and the shared library of skills, connectors, and agents; set the standards for agent and connector architecture so teams compose rather than rebuild.

Establish the target SDLC and common AI code / delivery tooling; define golden paths and the guardrails, gates, and review patterns that make AI-assisted and agentic development safe at bank-grade standards.

Evaluation, quality & security

Own eval and LLMOps infrastructure — versioning, regression and faithfulness testing — as the gate for shipping AI capabilities.

Embed AI security, data-classification enforcement, audit / logging, and use-case risk review at the platform layer, in close partnership with Information Security and governance.

Enablement & adoption

Own the tooling-access strategy and the enablement engine — onboarding, office hours, and change management — so adoption is measured and real, not nominal. The unlock is a real result on day one, not training.

Partner with the Product Operations and other teams to keep practices current and capture institutional knowledge before it is lost.

Organization & leadership

Build, develop, and lead a high‑performing organization of delivery and operational leaders, fostering accountability, continuous learning,