Executive Director, Agent Platform Lead

Executive Director, Agent Platform Lead

New York, NY, United States and 7 more

Apply Now

Job Information

Job Identification

210789551

Job Category

Client Solutions

Business Unit

Corporate Sector

Posting Date

11/09/2026, 08:45 PM

Locations

277 Park Ave, New York, NY, 10172, US

10 S Dearborn St, Chicago, IL, 60603, US

880 Powder Mill Rd, Wilmington, DE, 19803, US

1301 2nd Ave, Seattle, WA, 98101, US

3203 Hanover St, Palo Alto, CA, 94304, US

8181 Communications Pkwy Bldg F, Plano, TX, 75024, US

1111 Polaris Pkwy, Columbus, OH, 43240, US

575 Washington Blvd, Jersey City, NJ, 07310, US

Job Schedule

Full time

Base Pay/Salary

New York,NY $166,250.00-$260,000.00; Jersey City,NJ $166,250.00-$260,000.00; Seattle,WA $166,250.00-$260,000.00; Palo Alto,CA $166,250.00-$260,000.00; Chicago,IL $147,250.00-$235,000.00

Job Description

AMDP (AIML Data Platform) is where production AI gets built at the firm. SmartSDK, Agent Studio, the RAG SDK, the MCP SDK and the managed GenAI gateway are the paved paths our Lines of Business use to put agents into production. The AI/ML Product Specialist team is the applied side of that platform with engineers who sit next to LoB (Line of Business) teams and get hard use cases from pilot to production. You will work with teams in NYC, London, Hyderabad and Singapore, on use cases that touch every part of the firm.

You contribute towards agent we co-build with the firm's LoBs (Line of Business) and you decide what good looks like.

As Executive Director, Agent Platform Lead, you are the senior engineering authority for that builder’s side. You decide how production agents get built on AMDP, you set the technical bar for the Product Specialists shipping them, and you are the senior technical counterpart to the Agents & Evals Engineering organization that owns the underlying platform. This is a senior individual contributor role with no direct reports. Your influence comes from technical depth, architectural judgment, and being in the code and in the room at the same time.

You will personally architect the hardest engagements, review the code that ships under our name, and turn what we learn in the field into platform capability that Engineering then builds. You are the senior builder, who leverages architecture, code review, and the standards and governance to take agents into production. The platform gets better because you pushed for it. Governance, security, and evaluation are engineered into the paved paths not bolted on at review time. Your field learning goes directly onto the Agents & Evals roadmap.

We are builders first, we ship on paved paths, we write things down, we review each other's code, we take governance, security and evaluation seriously because our users do real work with real stakes. If that sounds like your kind of team, we would like to hear from you.

Job Responsibilities

Define opinionated, production-proven patterns across SmartSDK, Agent Studio, the RAG SDK and the MCP SDK for multi-agent orchestration, tool calling, MCP server design, state and memory management, human oversight hooks, and model access through the managed GenAI gateway.

Run architecture and code review for every agent we co-build or deliver. Mentor Product Specialists across NYC, London, Hyderabad and Singapore.

Embed with LoB (Line of Business) engineering teams on the highest complexity use cases. Write code where it matters. Get them to production on paved paths, not custom one-offs.

Partner with Agents & Evals product and engineering leaders on roadmap. Convert field signal into a ranked, evidence-backed list of capability gaps. Get Product Specialists into private preview so we know features before LoBs (Line of Business) do.

Work with Governance, Model Risk Management and Compliance so controls of the AgentSDK agent manifest, deployment gates, audit trail show up as engineering defaults, not late-stage review findings.

Define what “good enough to ship” means for our agents: onboarding gates, regression suites, tool-use fidelity, safety boundary tests, and shadow / canary evaluation in production.

Blast-radius containment, prompt injection mitigation, least-privilege tool access, and tool-surface review for every MCP integration we ship.

Maintain the Agent Developer Playbook, AI/ML paved paths and reference architectures to enable the platform scale and drive adoption of the platform.

Required Qualifications, Capabilities and Skills

10+ years in technology, with 3+ years building LLM-based applications or agent systems that ran in production.

Hands-on agentic system design: multi-agent orchestration, tool calling, MCP integration, RAG pipelines, context and memory management and a clear-eyed view of how these systems fail in production.

Working knowledge of running LLMs at scale: Evals, RAG, Harness Agents, Model Tuning and Serving, Rate limits, Model gateway patterns, and runtime evaluation.

Proven ability to set technical direction and raise engineering quality across teams without direct line authority through standards, reference architectures and review.

A track record of influencing a platform or product roadmap you did not own, with evidence that the capability you pushed for got built and used.

Writing and speaking that holds up with both principal engineers and senior executives.

Preferred Qualifications, Capabilities and Skills

MCP at scale: tool schema design, server architecture, tool-surface security and enterprise system integration.

Financial services, healthcare or government experience — including model risk management and audit-driven engineering.

Built or run agent evaluation frameworks: structured output evaluation, tool-use fidelity, safety boundary testing and production regression tracking.

Human-oversight engineering for autonomous systems: escalation paths, intervention hooks and auditability.

Breadth across the ML platform: model training, fine-tuning and serving, and data sourcing from Databricks, Snowflake and S3.

MS or PhD in Computer Science, Engineering or a related field or equivalent practical depth.

Production-grade Python is a plus.