Director, AI Architecture

Job Description

Scroll down the page to see all associated job requirements, and any responsibilities successful candidates can expect.

Position Summary

TKO Group Holdings, the parent of UFC, WWE, IMG, On Location and PBR, is deploying AI across multiple business units and a growing number of platforms.

The Director, AI Architecture & Agent Platform sets TKO's technical architecture for agents and AI platforms. This role determines how agents are designed, connected, evaluated, secured, observed and operated across TKO. It owns the technical point of view on orchestration, reusability, interoperability and platform choice, including where TKO should build, buy, use a platform-native capability or use an independent layer.

The role validates the architecture through targeted reference builds, including the first cross-system agent taken to production, then enables engineering and business-unit teams to deliver against reusable standards. It is not a central application delivery role or a high-volume build service.

This is an architecture-led individual-contributor role. It requires enough hands-on depth to prototype critical paths and challenge technical claims, along with the judgment and influence to guide work through teams that do not report to this role.

Responsibilities

1. Agentic Harness and Orchestration Architecture

• Own TKO's technical point of view on where orchestration and reusability should sit, including build versus buy and platform-native versus independent options.
• Define what each option requires, what it forecloses, and how agents built on different platforms will interoperate.
• Lead the TKO-side answer to the initial orchestration assessment, working with an external partner while challenging the analysis and making the recommendation.
• Define portability and migration criteria for platform-specific agents, including how TKO separates business logic, prompts, tools, identity, data access and evaluation from any one platform. Establish a path to rehost or retire agents built on IBM Watson, Salesforce or other platforms when warranted.
• Establish the architecture principles and decision records that keep the agent platform coherent as the estate grows.

2. Reference Architecture and Technical Validation

• Lead the architecture and technical direction for the first cross-system agent built through the new standard and taken to production.
• Personally prototype or implement the critical parts needed to validate the architecture, then partner with delivery teams on the broader build.
• Produce the templates, reusable components and engineering patterns that later builds will use.
• Make the first-of-a-kind patterns real without becoming the long-term application owner or central delivery team for every use case.

3. Technical Evaluation, Estimation and Portfolio Input

• Assess AI tooling and proposed solutions on technical merit at a level that can withstand challenge from engineers and vendors.
• Provide architecture, feasibility, effort, total-cost and technical-risk input into portfolio prioritization.
• Write technical assessments that make vendor claims, dependencies, limitations and tradeoffs explicit.
• Identify when a capability should be adopted, piloted, built, deferred or retired.

4. AI Engineering Standards, Guardrails and Sandboxes

• Define what must be true before an agent reaches production, including identity and access, evaluation and regression testing, observability, cost controls and failure handling.
• Define standards for agents that take consequential actions, including payments or writes to source systems: authorization, human approval, segregation of duties, transaction limits, audit trails, idempotency, rollback and failure handling, data protection and testing. Own the AI architecture standard while platform, payment, security and business owners retain responsibility for their underlying controls.
• Design and operate controlled sandboxes for emerging models and tools, including video-gen models such as Seedance. Give teams a safe way to test approved use cases without exposing TKO data, talent likeness, intellectual property or production systems.
• Advise Engineering teams adopting tools such as Claude Code on repository and data access, secrets, code review, testing, dependency and intellectual-property controls, logging, cost limits and approved use.
• Serve as the technical counterpart to Cybersecurity and other functions adopting AI tools. Help assess proposed uses, define safe operating practices, capture reusable patterns and align local decisions with TKO's enterprise architecture and guardrails.
• Set the access controls, logging, spend limits, data and content rules, evaluation requirements and path to production or shutdown for those