Senior AI/ML Engineer - Agentic Healthcare AI

Dillon Rosin

Senior AI/ML Engineer

+1-813-***-**** **************@*******.*** Gainesville, FL PROFESSIONAL SUMMARY

Senior AI/ML Engineer with 12+ years of experience in healthcare, progressing from data analytics and engineering to ML and agentic AI platforms across payer, provider, and digital health environments. Experienced in building production ML and LLM-based systems with a strong focus on HIPAA compliance, reliability, and measurable business impact. Hands-on leader with expertise in Python, SQL, AWS, LangGraph, LangChain, and MLOps/CI/CD, delivering scalable AI solutions through close collaboration with clinical, product, and compliance teams.

TECHNICAL EXPERTISE

Programming Languages Python, SQL, JavaScript, TypeScript, Bash/Shell, R Machine Learning & AI

Frameworks

PyTorch, TensorFlow, Keras, Scikit-learn, XGBoost, SHAP, LIME, LangGraph, LangChain, MCP, Django, FastAPI, React, Node.js

ML Engineering & MLOps NLP, BERT-based models, text classification, PHI detection, intent analysis, ETL, ELT, feature engineering, CI/CD, model monitoring

Data & Cloud Hugging Face, Apache Airflow, Apache Spark, Git, Dask, AWS S3, EMR, SageMaker, Lambda, Bedrock, Redshift, GCP BigQuery, Dataflow, Vertex AI, Docker, Kubernetes, Jupyter, PostgreSQL, Tableau, Power BI PROFESSIONAL EXPERIENCE

Principal Applied AI & ML Engineer BetterHelp Aug 2022 - Present

• Architected agentic AI systems composed of 8-12 task-specialized agents for intake analysis, risk scoring, therapist matching, and engagement optimization, serving 1M+ users, improving first-session satisfaction by 22-25%, and reducing early churn by 18-20%.

• Orchestrated end-to-end CI/CD pipelines for ML and agent workflows using Git-based versioning, testing, and staged deployments, supporting 30+ models and agent policies, reducing production release cycles from weeks to under 48 hours while maintaining 99.9% availability.

• Created a production-grade MLOps and AgentOps platform on AWS SageMaker, Feature Store, Lambda, and Step Functions, enabling automated model training, validation, canary releases, and rollback, cutting deployment failures by 60% and increasing safe experiment velocity by 45%.

• Established governance, monitoring, and compliance automation across 100% of AI agents and ML models, including drift detection, explainability checks, audit logging, and human-in-the-loop approvals, supporting HIPAA audits, reducing incident response time by 30%, and enabling weekly agent and model updates without downtime.

• Developed behavioral risk and engagement prediction models across 200+ behavioral signals, improving at-risk user identification accuracy by 15-18% and increasing outreach effectiveness by 15%. Sr. ML Architect TherapyNotes May 2019 - Jul 2022

• Architected and owned a production ML platform supporting 6+ predictive models across 300K+ patients, 5K+ providers, and 20M+ clinical records, achieving 99.9% pipeline reliability and enabling sub-2s inference latency for real-time clinical insights.

• Crafted end-to-end automated ML pipelines for 15+ EHR data sources, including structured clinical data, notes, and assessments, enabling near-real-time scoring and daily batch inference on 10M+ records.

• Implemented explainable ML frameworks using SHAP and LIME across 100+ engineered clinical features, increasing clinician adoption of ML insights by 40% and reducing model override rates by 30%.

• Optimized data ingestion, feature engineering, and model training workflows, cutting training time by 35-45%, lowering cloud compute costs by 20%, and increasing model refresh frequency from monthly to weekly.

• Led performance and cost optimization of Spark- and Airflow-based pipelines running 10M+ daily inferences, reducing model training time by 45%, lowering cloud costs by 20-25%, and improving ML reliability and data freshness SLAs. Sr. Data Engineer Optum Nov 2016 – Apr 2019

• Designed and maintained 20+ scalable data pipelines processing 5-8B+ healthcare records annually across claims, lab results, and provider data from 50+ source systems, supporting population health, utilization forecasting, and cost-prediction models.

• Transformed legacy BI analytics into ML-powered risk stratification models across 10M+ covered lives, improving high-risk patient identification accuracy by 22-28% and enabling earlier care interventions.

• Architected feature stores and reusable analytical datasets with 200+ engineered features, reducing data scientist experimentation time by 40% and accelerating model deployment cycles from weeks to days.

• Optimized Spark-based ETL and ML preparation workflows running on 100+ node clusters, cutting pipeline execution time by 30-35%, lowering compute costs by 25%, and improving downstream ML reliability and data freshness SLAs. Sr. Data Analyst Humana Jun 2013 – Oct 2016

• Delivered 30+ executive and operational dashboards supporting care quality, utilization management, and cost optimization initiatives covering 1M+ member records.

• Conducted statistical analysis and cohort studies across claims and clinical datasets, identifying utilization and outcome trends that informed

$8-12M annual cost reduction programs.

• Automated recurring reporting pipelines using SQL and Python, reducing manual effort by 50-60%, improving data accuracy by 35%, and cutting reporting turnaround from days to hours.

• Collaborated with clinical, actuarial, and operations teams on 20+ analytical initiatives, translating complex healthcare questions into measurable KPIs and laying the groundwork for predictive analytics and ML adoption. EDUCATION

University of Florida Gainesville, FL

Bachelor's Degree in Computer Science 2009 - 2013

Contact this candidate