Malik Warren

login.gov (GSA / TTS) and U.S. Air Force programs · Jul 2022 – present

• Core contributor to the identity-verification (IDV) pipeline on login.gov, the federal NIST 800-63 IAL2 identity-proofing platform serving 100M+ user accounts across 50+ agencies. 46 merged pull requests over roughly 13 months in the proofing and access-control layer.
• Trusted peer reviewer on an access-critical federal codebase: reviewed 64 colleagues' pull requests and formally approved 53 over the same period.
• Built AAMVA (DMV) identity-verification gating that blocks applicants from advancing until their state-ID data passes an authoritative-source check. Led the AAMVA reimplementation, a 1,438-line change re-architecting the flow with pending-session state and async DMV polling.
• Designed a feature-flag-gated Proofing Agent API with mandatory request-header authentication, returning 404 when the capability is disabled and 400 on missing identifiers. Grants and scopes access to a sensitive capability based on approved use.
• Hardened passport MRZ validation, adding Department of State verification, expiration enforcement, and US-issuer restriction as gates controlling whether a proofing session may proceed.
• Identified and patched a file-upload exploit in the Acuant document-capture SDK across all deployed versions. Built the LexisNexis TrueID (DDP) document-authentication client and a manual-capture fallback, improving completion rates.
• Enforced safeguards across the flow: gated downstream vendor calls on a passing source check, closed a fraud-check bypass for hybrid-flow users missing ThreatMetrix sessions, and added anti-bypass flow-policy enforcement.
• On U.S. Air Force programs: core contributor to 5 production Go microservices on AWS GovCloud. Built the JWT authentication infrastructure (HMAC signing, token encrypt/decrypt, Secrets Manager), SAML 2.0 SSO, custom CSRF middleware, and reusable session/permission middleware extracted into a shared module.

Cleared federal AI delivery · Jan 2026 – present

• Led design, development, and deployment of a cleared AI system for a federal agency, managing a 6-person team across 5 sprints.
• Designed the full LLM architecture: an XML-structured prompt system (~6K tokens), full-context injection, forced tool use for structured JSON via Bedrock's toolChoice API, and chain-of-thought reasoning. Iterated accuracy from a 67% baseline to 98%, with 0% underclassification.
• Built a provider abstraction layer (circuit breaker, exponential backoff, fallback chains) supporting Anthropic direct, AWS Bedrock, and Ollama behind a single factory interface, plus a benchmarking framework (14 metrics, A/B prompt testing, severity-weighted safety scoring).

Government technology consultancy · Jan 2022 – 2025

• Selected for Hutch, the two-year government-contracting incubator, as one of eight companies in the Class of 2024.
• Delivered $750K+ in lifetime contracts as a Google Cloud Partner, with clients including the U.S. Census Bureau, Harvard Business School, Morgan State University, and Texas Southern University.
• Built the ByBlack Spend Tracker (React, Python, MongoDB, GCP) with Plaid integration, and MyBlockCounts with the University of Maryland, a React Native geospatial civic-reporting tool routing resident feedback to government agencies.
• Developed federal environmental-justice spending visualizations (React, Node.js, PostgreSQL, GCP) with multiple HBCUs and the Bezos Earth Fund.

Dec 2021 – Jul 2022

• Led SummitRTS, a cloud-automation framework with GCP integration for provisioning, testing, and security. Built the WorkerManager provisioning system (Node.js, Express, PostgreSQL) and its React / GraphQL frontend.

Jul 2020 – Dec 2021

• Worked on the Defense Travel System, a mission-critical app serving 100K+ military and DoD personnel daily (Java, Spring Boot, SQL). Built internal hiring/clearance-tracking and resource-allocation tools.

Jan 2020 – Jan 2022

• Built a client onboarding and tracking app (Django, PostgreSQL, JavaScript) replacing a manual intake process, with reporting that supported grant proposals.