Richard Young, Ph.D.

Senior AI Research Scientist

UHG Enterprise IRB Co-Chair | UNLV Professor | AI Safety & Clinical AI

Applying foundation models and large-scale evaluation methods to high-stakes domains: AI safety, cybersecurity, clinical care, neuroscience research acceleration, and decision-support systems.

Senior AI Research Scientist, UnitedHealth GroupCo-Chair, UHG Enterprise IRBPart-time Professor, UNLV Lee Business SchoolKeynote speaker, John Snow Labs Applied AI Healthcare

Connect on LinkedIn View Resume Academic CV

Professional Profiles

LinkedIn GitHub Ollama HuggingFace Scholar ORCID

110M+

Lives Under IRB Scope

UHG, Optum, Reliant, affiliates

27+

Publications

AI safety, clinical AI, neuroscience

80+

Open Model Releases

Hugging Face and Ollama

97K+

Adversarial API Queries

TEMPEST across 10 frontier models

About

Richard Young is an AI researcher operating at the intersection of foundation-model evaluation and consequential applied domains. His work empirically measures how frontier models fail under adversarial, clinical, and instruction-following pressure, then ships the benchmarks, datasets, and models needed to harden them.

He serves as Senior AI Research Scientist at UnitedHealth Group, Co-Chair of the UHG Enterprise IRB covering UnitedHealth Group, Optum, Reliant Medical, and international affiliates, and part-time Professor at the UNLV Lee Business School.

The path forward lies in creating partnerships that think with us, amplifying human potential while preserving our values.
Dr. Richard Young

With over a decade of experience at the intersection of neuroscience, machine learning, and clinical research, Dr. Young's work demonstrates that the greatest challenges in healthcare, scientific discovery, and human understanding become tractable when we pair human creativity and domain expertise with the computational power of modern AI systems.

This research-to-artifact loop spans peer-reviewed publications, open datasets, open-weight model releases on Hugging Face and Ollama, and production agentic systems serving large-scale healthcare operations.

His current programs cover AI safety and alignment, code safety and cybersecurity, clinical privacy and fairness, neuroscience research acceleration, and decision-support systems.

Warm regards,

Richard Young, Ph.D.

Senior AI Research Scientist | UHG Enterprise IRB Co-Chair | UNLV Professor

Current Focus

Cybersecurity Code-Safety Program

Four-paper collaboration with Dr. G. D. Moody: validated 1,554-prompt malicious-code prompt bank, systematic review of thirteen prompt corpora, 10-model refusal benchmark, and mechanistic abliteration study.

arXiv:2605.03179USENIX Security targetNeurIPS target

Reasoning-Model Safety Series

Active research line on chain-of-thought faithfulness, inverse scaling at test-time compute, sandbagging detection, internal hallucination probes, unlearning, MCP protocol safety, and steganographic reasoning.

19-paper seriesopen-weight auditsreasoning models

Clinical AI & Healthcare Deployment

Parkinson’s biomarker extraction, EQUITRIAGE gender-bias audits, MIMIC readmission prediction, ICD prediction, clinical note summarization, ED wait-time prediction, and post-arrest prognosis.

clinical AIfairnessprivacy

Production AI Work

Conversational analytics, RAG, and agentic systems running on Databricks.

CxS Insights

53 tools

Production Databricks App serving Optum/UHC Customer Experience Services with 47.5M member records and 26.4M vector-indexed call transcripts.

Ask Lucky

85 tools

Plain-English analytics over 41.5M+ member profiler records, NOC data, drug datasets, and risk/intervention workflows.

Ask Thaur

60x faster

Conversational SQL over 150M+ call-center records, reducing two-hour analyst workflows to 15-35 second answers.

Ask Richard

56-book corpus

Multimodal LLM and RAG training platform for 80+ data scientists, indexing 27,982 documents and 6,687 images.

Skills & Technologies

Foundation Model Evaluation

Adversarial TestingRefusal EvaluationCoT FaithfulnessInstruction FollowingSafety Guardrails

Clinical AI & Governance

Enterprise IRB OversightPHI Risk EvaluationED Triage FairnessClinical Decision SupportProtocol Review

Cybersecurity Research

Code-Safety BenchmarksPrompt Bank DesignMalicious-Code CorporaAbliteration StudiesMechanistic Audits

Production AI Systems

Databricks AppsAgentic Tool RoutingVector SearchSSE StreamingCost-Aware Model Routing

Open Model Ecosystem

Hugging FaceOllamaGGUFMLX QuantizationEmbedding Models

Teaching & Leadership

Graduate ML InstructionHealthcare AI CurriculumKeynotesResearch MentorshipTechnical Writing

Work in Progress

Books, preprints, and ongoing research projects

Book in Progress2026

The Neuroscience of Artificial Intelligence

Young, R.J.

A 38-section NeuroAI Handbook integrating neuroscience, mathematical frameworks, modern AI systems, and 100+ research-paper deep dives.

Explore the book

Graduate Textbook in Progress2026

Healthcare Analytics and AI: Building Systems That Actually Work

Young, R.J.

A 22-chapter, 471-page graduate text covering healthcare data pipelines, clinical NLP, LLMs, agentic workflows, fairness, privacy, and deployment governance.

Explore the book

Patents & Innovations

Translating research into clinical practice

US Patent ApplicationUS 20240395417 A1

Systems and Methods for Determining Readmission Rates

Inventors: Gada, E., Young, R.J. • Assignee: Optum Inc.

Filing Date:May 25, 2023

Publication Date:November 28, 2024

Novel methods for obtaining hospital admission data, determining primary admission and readmission values for disease-specific groups, calculating disease-specific readmission rates, and presenting actionable insights via user interfaces.

View on Google Patents

Teaching

Educating the next generation of AI practitioners

Business Intelligence & Machine Learning

UNLV Lee Business School

•

Part-Time Instructor

•

2023 - Present

Graduate-level courses focusing on practical ML implementation, transformer architectures, and ethical AI frameworks.

Large Language ModelsDeep LearningBusiness AnalyticsAI EthicsModel Deployment

Speaking Engagements

Delivered 35+ invited talks to audiences of 250-50,000 on:

Large Language ModelsAI SafetyClinical AIAI in Education

View All Teaching Materials

Research Focus

AI Safety & Alignment Evaluation

Large-scale adversarial evaluation of frontier and open-weight models, including TEMPEST replication, instruction-following failures, guardrail robustness, and chain-of-thought faithfulness.

Code Safety & Cybersecurity

A four-paper code-safety research program with Dr. G. D. Moody measuring malicious-code refusal behavior, prompt-bank validity, and separability of code-safety directions in activation space.

Clinical AI at Deployment Scale

Production agentic systems, clinical privacy audits, ED triage fairness, clinical note workflows, and scientific oversight for research protocols across enterprise healthcare populations.