Senior Systems Engineer · Corporate Systems Lead, Robinhood
Identity is what I do: who gets in, what they can touch, and how access ends when they leave. Ten years of building that across fintech and healthcare: sign-in and provisioning, governance and audit, acquired companies folded into one standard, and AI platforms: accounts, scopes, and an offboard like everyone else. The question never changes: who, or what, is asking, what may it do, and can I prove it later.
systems in scope · hover or click one
Okta, Google Cloud, and Google Workspace at a public fintech: every login, permission, and offboarding runs through these. Green rows are in production and audited; amber rows are being built or hardened right now.
Authentication in Okta: SAML, OAuth 2.0, OIDC, and SCIM to everything downstream. The whole lifecycle from Workday: birthright access, joiners, movers, leavers, rehires, service accounts, and the edge cases SCIM can't reach, closed with Okta Workflows and Python.
Led the Okta Identity Governance rollout: access certification campaigns and policy-driven lifecycle controls. Audit responses across SOX controls, access reviews, and service accounts, working directly with external auditors. Identity changes ship through technical reviews I author.
Intake for acquired companies: their apps and their identity provider brought onto the Okta standard as one repeatable process, not a per-deal scramble. Standalone IdPs are retired once their people are across; systems we already run are merged into ours.
Claude, Cursor, ChatGPT, and Gemini Enterprise get accounts, SCIM, roles, and an offboard like everyone else. An MCP server touching Jira or Slack acts as itself, not as whoever installed it. Service-account credentials are checked out of the vault at runtime and returned. Identity, not model safety.
Google Cloud IAM and project structure as Terraform, so engineers get a paved road from prototype to production instead of asking for exceptions.
Google Workspace and Slack, where every employee and every new AI platform lands first. Who can share what outside the company, which third-party apps may connect, and what lands in the audit trail.
Ten years of identity work as a Terraform plan: + is something that exists because I built it, ~ something I changed, - something I removed. Every line is on the .
Three things in motion, each with the artifact or the shape behind it.
Okta's groups, group rules, apps, and sign-on policies live in Terraform, applied from CI with the plan attached to every change. Drift shows up in a diff, not in an audit. The foundation is public: state split per stack, so identity, apps, policies, authorization, and governance can be owned by different teams without sharing one state file.
A paved road from local prototype to hosted service. Project factories, IAM bindings, and access controls all live in Terraform, so shipping an AI-assisted tool to production is a reviewed change, not a hand-built exception.
Tightening Google Workspace: access policies, DLP, third-party OAuth, audit coverage. The attack surface gets bigger every time someone connects a new AI platform, and that's the part I'm watching.