Sean Dyer
Leawood
Summary
Senior Data Scientist leveraging 20+ years in machine learning to drive business solutions. Specializes in developing production models for cybersecurity and fraud detection, with a proven ability to implement LLM validation and generative AI systems that enhance operational security and business value.
Overview
24
24
years of professional experience
1
1
Certification
Work History
Associate Director, AI/ML Cybersecurity Practice
Ascendion
Bedminster
01.2023 - Current
- LLM Validation & AI Safety (Major Financial Institution): Led comprehensive adversarial testing initiative to validate LLM-based systems for production deployment. Developed rigorous evaluation metrics and adversarial prompting techniques to systematically test for hallucinations, ungrounded responses, toxic content generation, and other failure modes. Implemented LLM-as-a-judge methodology to quantitatively measure validation metrics at scale. Provided actionable feedback enabling client teams to refine guardrails, content filters, and safety controls, ensuring secure enterprise deployment of generative AI systems.
- Built and deployed generative AI models for Fortune 100 clients using LangChain, LangGraph, and AWS Bedrock, enhancing client capabilities in AI-driven solutions.
- Design and implement custom ML solutions addressing cybersecurity, fraud detection, and operational analytics.
- Developed multi-agent AI systems integrating tools for automated workflows and decision support, streamlining operations and improving decision-making processes.
- Led team of data scientists and ML engineers on complex projects, driving innovation and ensuring successful project delivery.
- Build proof-of-concept models demonstrating gen AI capabilities and business value.
AI Architect (Data Science Focus)
Pearson (via Ascendion)
01.2023 - 12.2023
- Developed ML models for automated PI/SI detection and redaction ensuring GDPR/FERPA compliance.
- Created features from complex educational datasets to enhance performance of downstream ML applications.
- Analyzed educational platform data to identify patterns and insights that informed AI strategy.
- Worked with engineers and product teams to implement data science solutions.
Lead Data Scientist
AT&T (MSSP - now LevelBlue)
01.2022 - 12.2023
- Developed and deployed ML models for threat detection, anomaly detection, and security analytics on AWS SageMaker, enhancing security posture.
- Created end-to-end ML pipelines from data ingestion through model deployment and monitoring, streamlining workflow and improving model management.
- Improved model accuracy by 25% through feature engineering and algorithm selection.
- Built production LLM applications using AWS Bedrock for automated threat intelligence analysis and incident response.
- Guided data scientists in model development, experimentation, and deployment best practices, fostering knowledge sharing and improving team capabilities.
Principal Scientist
SAS Institute
Cary
01.2021 - 12.2022
- Translated academic research into production-ready forecasting algorithms, improving deployment speed and model reliability.
- Conducted rigorous model evaluation and comparison studies, identifying optimal techniques for forecasting accuracy.
- Implemented novel forecasting approaches combining classical statistics with modern ML.
- Developed advanced statistical and ML forecasting models to enhance accuracy of time-series predictions.
Data Science Manager
SAS Institute - SAS Talon Product
Cary
01.2013 - 12.2021
- Implemented LSTM neural networks for domain generation algorithm (DGA) detection achieving 95%+ accuracy.
- Developed unsupervised learning models for identifying malicious network behavior and insider threats.
- Built ML models for network threat detection, user behavior analysis, and risk scoring.
- Created peer-based behavioral models for risk assessment, contributing to patented innovations (US Patent #10,498,755 B2).
- Deployed models processing millions of network events per day with sub-second latency.
- Designed 100+ features from raw network data for threat detection models.
- Built ML pipeline integrating Python with high-performance analytics platform, enhancing model deployment efficiency.
- Hired and mentored data science team, establishing coding standards and review processes to improve collaboration and code quality.
Technical Architect (Data Science)
SAS Institute - US Federal Sales
Cary
01.2005 - 12.2013
- Delivered data science solutions for federal agencies, enhancing decision-making in forecasting, text analytics, and operations research.
- Built custom ML models for classification, regression, clustering, and optimization problems.
- Developed proof-of-concept models to showcase ML capabilities, facilitating client understanding and adoption of data-driven approaches.
- Trained government analysts on ML techniques and statistical methods, empowering them to leverage advanced analytics in their projects.
- Performed advanced statistical analysis and hypothesis testing for government clients.
Program Manager / Data Science Lead
Sprint
Overland Park
01.2002 - 12.2005
- Managed analytics for operations generating $400M+ annual revenue.
- Built classification models identifying high-risk applications with 80%+ precision.
- Developed ML models for point-of-application fraud detection, minimizing wireless service losses.
- Performed exploratory data analysis on customer application data to identify fraud patterns.
- Assessed vendor models and statistical methods, ensuring alignment with fraud detection goals.
Education
Master of Arts - Mathematics/Statistics
Southern Illinois University
Carbondale, ILBachelor of Science - Mathematics
Southern Illinois University
Carbondale, ILSkills
Machine Learning
- Deep Learning
- PyTorch
- TensorFlow
- Scikit-learn
- XGBoost
- LightGBM
- Neural networks
- LSTM
- CNN
- Transformers
- Ensemble methods
- Gradient boosting
- Regularization
- Anomaly detection
- Behavioral analytics
- Time-series forecasting
- NLP
- Computer vision
- Feature engineering
- Hyperparameter tuning
- Model selection
- Cross-validation
- A/B testing
- Generative AI
- LLMs
- LangChain
- AWS Bedrock
- OpenAI API
- Hugging Face Transformers
- Prompt engineering
- RAG
- Fine-tuning
- Agentic AI
- Adversarial testing
- Hallucination detection
- LLM as a judge
- Guardrail development
- LLM integration
- Chatbots
- Automated insights
- Document processing
- Data Science
- Analytics
- Python
- NumPy
- Pandas
- Matplotlib
- Seaborn
- Plotly
- R
- SQL
- Hypothesis testing
- Regression analysis
- Bayesian methods
- Experimental design
- Apache Spark
- Distributed computing
- Streaming analytics
- Jupyter
- Git
- Docker
- VS Code
- MLflow
- AWS SageMaker
- Azure ML
- Google Cloud Platform
- Model serving
- Containerization
- CI/CD pipelines
- Monitoring
- SQL/NoSQL databases
- Data lakes
- Feature stores
- ETL pipelines
Websites
Certification
- NVIDIA-Certified Professional: Accelerated Data Science, 05/01/25, 05/01/27
- NVIDIA-Certified Associate: AI Infrastructure and Operations, 12/01/24, 12/01/26
- NVIDIA-Certified Associate: Generative AI LLMs, 12/01/24, 12/01/26
- NVIDIA-Certified Associate: Generative AI Multimodal, 12/01/24
- US Department of Defense Top Secret Security Clearance, Inactive
Accomplishments
- LLM Safety Leadership: Led enterprise-scale LLM validation initiative for major financial institution, developing comprehensive adversarial testing framework.
- Production ML: Deployed 20+ ML models into production environments processing billions of events.
- Model Performance: Achieved 95%+ accuracy on DGA detection using LSTM deep learning.
- Innovation: Named inventor on 3 US patents for novel ML techniques in cybersecurity.
- Gen AI Adoption: Early adopter of LLM technology, deployed production LLM applications in 2023.
- Business Impact: Models directly supported $400M+ in revenue and prevented millions in fraud losses.
- Team Leadership: Built and led data science teams from startup through product maturity.
Technical Skills Keywords
Python, R, SQL, Bash, PyTorch, TensorFlow, Scikit-learn, XGBoost, LightGBM, Keras, ONNX, NumPy, Pandas, Matplotlib, Seaborn, Plotly, SciPy, Statsmodels, LangChain, LangGraph, AWS Bedrock, OpenAI API, Hugging Face Transformers, prompt engineering, RAG, fine-tuning, Adversarial testing, red teaming, hallucination detection, LLM-as-a-judge, guardrails, content filtering, AI safety, LSTM, CNN, Transformers, neural networks, backpropagation, gradient descent, Supervised learning, unsupervised learning, reinforcement learning, ensemble methods, random forests, gradient boosting, SVM, decision trees, Regression analysis, hypothesis testing, Bayesian statistics, time-series analysis, experimental design, A/B testing, AWS SageMaker, AWS Bedrock, Azure ML, GCP AI Platform, AWS EC2, AWS S3, AWS Lambda, Apache Spark, PySpark, distributed computing, streaming analytics, Kafka, Docker, Kubernetes, MLflow, Git, GitHub, CI/CD, model monitoring, model serving, PostgreSQL, MySQL, MongoDB, Cassandra, SQL, NoSQL, data lakes, ETL, Jupyter Notebooks, VS Code, Git, GitHub, JIRA, Confluence, Cybersecurity, fraud detection, anomaly detection, forecasting, NLP, computer vision, behavioral analytics, LLM validation, AI safety, Agile, Scrum, feature engineering, hyperparameter tuning, cross-validation, model selection, model evaluation, adversarial testing, US Department of Defense Top Secret (Inactive)
Publications
- Multi-Method Ensemble Approach to AI-Generated Text Detection, LinkedIn Article, 2024, https://www.linkedin.com/pulse/multi-method-ensemble-approach-ai-generated-text-detection-sean-dyer-iuc6c/
- Evaluating Small AI Models for Hallucination Detection, LinkedIn Article, 2024, https://www.linkedin.com/pulse/evaluating-small-ai-models-hallucination-detection-sean-dyer-axotc/
Patents
- US Patent #10,498,755 B2, Cybersecurity Risk Score Computation Based on Peer Grouping, 2019
- US Patent #10,841,326 B2, Cyber Security System, 2020
- US Patent #10,965,706 B2, Cyber Security System, 2021
Core Technical Skills
PyTorch, TensorFlow, Scikit-learn, XGBoost, LightGBM, Keras, Neural networks (LSTM, CNN, Transformers), ensemble methods, gradient boosting, regularization, Anomaly detection, behavioral analytics, time-series forecasting, NLP, computer vision, Feature engineering, hyperparameter tuning, model selection, cross-validation, A/B testing, LangChain, LangGraph, AWS Bedrock, OpenAI API, Hugging Face Transformers, Prompt engineering, RAG (Retrieval-Augmented Generation), fine-tuning, agentic AI, Adversarial testing, hallucination detection, LLM-as-a-judge, guardrail development, LLM integration, chatbots, automated analysis, document processing, Python (NumPy, Pandas, Matplotlib, Seaborn, Plotly), R, SQL, Hypothesis testing, regression analysis, Bayesian methods, experimental design, Apache Spark, distributed computing, streaming analytics, Jupyter, Git, Docker, VS Code, MLflow, AWS SageMaker, Azure ML, Google Cloud Platform, Model serving, containerization, CI/CD pipelines, monitoring, SQL/NoSQL databases, data lakes, feature stores, ETL pipelines
Timeline
Associate Director, AI/ML Cybersecurity Practice
Ascendion
01.2023 - Current
AI Architect (Data Science Focus)
Pearson (via Ascendion)
01.2023 - 12.2023
Lead Data Scientist
AT&T (MSSP - now LevelBlue)
01.2022 - 12.2023
Principal Scientist
SAS Institute
01.2021 - 12.2022
Data Science Manager
SAS Institute - SAS Talon Product
01.2013 - 12.2021
Technical Architect (Data Science)
SAS Institute - US Federal Sales
01.2005 - 12.2013
Program Manager / Data Science Lead
Sprint
01.2002 - 12.2005
Master of Arts - Mathematics/Statistics
Southern Illinois University
Bachelor of Science - Mathematics
Southern Illinois University