Yousuf Mohammad

Forward-Deployed AI Leader · UnitedHealth Group

Yousuf Mohammad

I take healthcare GenAI from idea to production, owning strategy, product, architecture, engineering and delivery end to end.

Portrait of Yousuf Mohammad
  • Richmond, Texas
  • 20+ Years
  • Previously Fannie Mae & AWS
  • Doctoral Researcher, UTHealth Houston
  • Claude Production AI Specialist · AWS Golden Jacket · CNCF Kubestronaut
Download Resume Contact

At a Glance

  • Forward-Deployed GenAI 3 weeks from idea to MVP for a Medicare Part D GenAI solution, embedded with business teams
  • Agentic AI 2× regulated cloud adoption with CLARA, a multi-agent AI system I designed and built
  • Enterprise GenAI $3M saved per year with LLM-powered document intelligence at Fannie Mae
  • AI-Ready Platforms $67M+ annual cloud spend made AI-ready across 160+ healthcare migrations

Focus Forward-Deployed AI · Generative & Agentic AI · Healthcare & Regulated Industries · AWS & Azure

Perspective

Most AI never makes it past the demo. It stalls in the gaps between strategy, product, architecture, engineering and operations, where every handoff loses context and no one owns the outcome end to end.

How I Close the Gap

  1. Value Before Models

    Find where AI creates measurable value, and where it doesn't, before writing a line of code.

  2. Own the Whole Loop

    Set strategy with business leaders and build alongside engineers. Fewer handoffs mean faster, safer delivery.

  3. Built for Production

    Evaluation, guardrails and compliance from day one. That is the difference between a pilot and a platform.

Selected Work

UnitedHealth Group

GenAI for Medicare Part D

Challenge
Identify the highest-impact drugs to inform the Medicare Part D bidding process.
Approach
Embedded with business teams, owning strategy, product, architecture and hands-on engineering of an evidence-based GenAI solution.

3 weeks

from idea to MVP; now being taken to production

UnitedHealth Group

CLARA: Agentic Cloud Governance

Challenge
Architecture governance was slowing regulated healthcare workloads on their way to the cloud.
Approach
Designed and built a multi-agent AI system that automates architecture reviews and risk scoring for regulated workloads.

StackAgentic AI · Multi-Agent Systems · Automated Risk Scoring

2×

regulated cloud adoption

Before
25%
After
50%

Fannie Mae

GenAI Document Intelligence

Challenge
Make a large estate of lending documents faster to find and review.
Approach
Re-architected document management as a plugin-based platform with metadata tagging, search and analytics, and LLM-powered summarization.

StackOpenAI API · Amazon S3 Tables · Amazon Athena · Plugin-Based CMS

$3M

saved per year, with 60% faster retrieval

Amazon Web Services

Enterprise Kubernetes Platform

Challenge
Standardize and speed up how enterprise teams build and deploy containerized workloads.
Approach
Built a global, enterprise Kubernetes framework, backed by reference architectures and proofs of concept.

StackKubernetes · AWS · Reference Architectures

30%

faster deployment cycles

Doctoral Research

Doctor of Health Informatics
UTHealth Houston

I've spent years on the systems behind care. Now I'm taking that rigor to the point of care.

NephroIQ: Nephrology Guidelines in Everyday Practice

NephroIQ is a platform for AI agents that support nephrologists and their patients. Its first agent, Filtra, prepares a one-page, evidence-backed summary before each chronic kidney disease (CKD) visit, built on KDIGO guidelines. It shows the evidence and its grade behind every point and leaves every decision with the nephrologist. It is being developed with a community nephrology practice in Beaumont, Texas.

Research Question Does a pre-visit evidence summary, compared with usual chart review, increase guideline-recommended treatment for adults with CKD over 12 weeks?

How Filtra Works

  1. Patient Chart Labs, notes and history from the health record
    Evidence Library KDIGO guidelines, approved by a nephrologist
  2. Filtra AI agent matches the chart to the evidence
  3. Pre-Visit Summary One page, with the evidence and grade behind every point
  4. Nephrologist Makes the call; the decision is checked against the guideline
Filtra prepares; the clinician decides. It never orders, prescribes or alerts.

Why It Matters

  • 14.8% of patients with CKD are on an SGLT2 inhibitor, a therapy guidelines now recommend for most adults with CKD Kovesdy et al., 2025 · KDIGO, 2024
  • 19 patients treated for about 2.4 years prevent one major kidney or heart event DAPA-CKD trial, NEJM 2020
  • 7.5 h in the health record for every 8 hours of patient care in nephrology Holmgren et al., 2024

Built to Be Trusted

  • Every point is traced to a nephrologist-approved evidence library. Before any pilot, offline testing must show zero fabricated claims and ≥95% consistency on repeat runs.
  • Designed to meet the FDA's criteria for non-device clinical decision support: it explains its basis, and never orders, prescribes or alerts.
  • De-identified data for testing; identified data only in an IRB-approved pilot, in a HIPAA-eligible environment.

Plan

  1. 2026–27 · nowIdentify the gap: evidence review and clinic workflow study
  2. 2027–28Build the evidence library and Filtra; offline testing and IRB review
  3. 2028–2912-week clinic pilot, evaluation and sustainment plan

Writing

All articles on Medium →

Experience

Trajectory

  1. 2002Software Engineer
  2. 2012Team Lead
  3. 2016Technology Manager
  4. 2019Technical Architect
  5. 2021Solutions Architect
  6. 2023Principal
  7. 2026Forward Deployed AI Lead

UnitedHealth Group

Sep 2025 – Present
Minnetonka, MN

  1. Forward Deployed AI Lead · Jul 2026 – Present

    Mandate expanded to AI strategy, product and delivery

    • Led the team that took a Medicare Part D GenAI solution from idea to MVP in 3 weeks; now driving it to production (see Selected Work).
    • Embedded with business teams, owning strategy, product, architecture, hands-on engineering and delivery end to end.
    • Claude Code Ambassador Program: certified Claude Production AI Specialist; helped 20+ colleagues put AI to work.

  2. Principal Architect, Medicare & Retirement · Sep 2025 – Jul 2026

    • Pioneered CLARA, a multi-agent AI system for architecture governance and risk scoring that helped double regulated cloud adoption from 25% to 50%.
    • Led the transformation of legacy healthcare platforms into AI-ready, cloud-native architectures: $67M+ annual spend, 160+ migrations.
    • Set technical direction for 11+ architects and 80+ engineers across multiple vendors; standardized Cloud & GenAI accelerators and enterprise Kubernetes platforms.

StackGenAI · Agentic AI · Azure · AWS · Kubernetes · HIPAA-Aligned Architecture

Fannie Mae

Aug 2023 – Jul 2025
Reston, VA

Principal, Enterprise Architecture & Technology · Director Level

  • Brought GenAI into document management with OpenAI-powered summarization and metadata extraction (see Selected Work).
  • Designed a domain-driven, API-first microservices architecture for the lending platform, reducing processing fees by 0.3%.
  • Replaced manual, PDF-based lender risk assessments with an event-driven, configuration-driven React forms engine that achieved 99% approval ratings.
  • Built a self-service analytics platform on Redshift APIs with Lambda-based pipelines, cutting development cycles by 40%.

StackOpenAI API · AWS Lambda · Redshift · Athena · S3 Tables · React · DDD · Event-Driven Architecture

Amazon Web Services

Apr 2021 – May 2023
Houston, TX

Solutions Architect

  • Drove $40M ARR (65% YoY) across North America, the UK and Germany, aligning Fortune 500 C-suites on cloud strategy through executive briefings.
  • Delivered 30% cost savings and faster disaster recovery through Well-Architected Framework reviews.
  • Built an enterprise Kubernetes platform and a React micro-apps framework that cut duplicate development by 40%.
  • Inducted into the AWS Technical Field Community; trained 300+ engineers through workshops and game days.

StackAWS · Kubernetes · Well-Architected · Micro Frontends · FinOps

Earlier Career

2002 – 2023

Principal Cloud Architect at Kessel Run (US Air Force); Solutions Architect at Corteva Agriscience; and Software Engineer to Technical Architect at Schlumberger, Shell, JPMorgan Chase, Wells Fargo and others, building secure, polyglot systems (Java, C#, Python, JavaScript) for regulated industries.

Technical Depth

Leading GenAI well means knowing what it takes to run it in production. So I still build, from LLM integration and evaluation to the cloud platforms underneath.

Generative AI

  • Agentic & Multi-Agent Systems
  • RAG
  • LLM Integration (OpenAI, Claude)
  • Evaluation & Guardrails
  • NLP & ML Pipelines
  • Explainable AI

Cloud Platforms

  • AWS & Azure Strategy
  • Kubernetes at Enterprise Scale
  • Serverless
  • CI/CD & IaC
  • Observability
  • FinOps & Cost Optimization

Architecture

  • Domain-Driven Design
  • Event-Driven, CQRS & Event Sourcing
  • API-First Microservices
  • Micro Frontends
  • Legacy Modernization (Strangler Pattern)
  • HIPAA-Aligned Compliance

Engineering

  • Python
  • Node.js
  • React
  • Blazor
  • Java / Spring Boot
  • .NET Core
  • Kafka
  • Amazon Redshift
  • Athena
  • Azure Data Lake

Credentials & Recognition

Certifications

Amazon Web Services

AWS Golden Jacket All 8 active AWS certifications
  • Solutions Architect Professional
  • DevOps Engineer Professional
  • Solutions Architect Associate
  • Data Engineer Associate
  • Developer Associate
  • SysOps (now CloudOps) Associate
  • AI Practitioner
  • Cloud Practitioner

Kubernetes

CNCF Kubestronaut All five CNCF Kubernetes certifications
  • Kubestronaut
  • CKA Administrator
  • CKAD Application Developer
  • CKS Security Specialist
  • KCNA Cloud Native Associate
  • KCSA Security Associate

Microsoft Azure

4 active certifications
  • Solutions Architect Expert
  • DevOps Engineer Expert
  • Administrator Associate
  • Azure Fundamentals

Generative AI

UnitedHealth Group
  • AI Dojo Generative AI
  • Claude Production AI Specialist

Architecture & Delivery

  • TOGAF® 9 Certified
  • Project Management Professional

Education

  • Doctor of Health Informatics UTHealth Houston, McWilliams School of Biomedical Informatics. In progress; translational project: NephroIQ
  • Master of Science Computer Science, University of Houston–Clear Lake
  • Bachelor of Science Computer Science, Osmania University

Recognition

  • 6× BEAT AwardFannie Mae
  • Solutions Architect of the QuarterAWS
  • CIO AwardTeam Contribution
  • SpeakerAWS Energy Conference, London
  • AuthorFour articles on forward-deployed AI and healthcare AI (see Writing)
  • VolunteerMicrosoft TEALS, teaching computer science in high schools

Open to conversations about taking healthcare AI from strategy to production.

yousuf.mohammad@gmail.com · Richmond, Texas

Email LinkedIn