Rochester, Minnesota
I am a software engineer at Mayo Clinic IT focused on cloud-native platforms and AI services. I design and operate secure backend services, agentic AI tooling, and GPU-ready compute environments that help teams build and deploy AI workloads reliably.
My work spans Terraform-based automation on GCP, secure backend services for AI access, SLURM-based GPU clusters, and observability for ML systems. I enjoy building platform capabilities that improve developer velocity while raising the reliability and safety bar by default.
Software Engineer, Generative AI Program · Mayo Clinic IT
M.Sc. Biomedical Engineering and Physiology · Mayo Clinic Graduate School
B.Sc. Industrial Engineering and Management · Ben Gurion University
The production capabilities I’ve designed, deployed, and operated.
This section focuses on outcomes and responsibilities rather than tools.
Designed and built core components of a Patient MCP Server — a FastAPI-based Model Context Protocol service that exposes structured clinical data to LLM-driven agents, with tool routing across multiple backends and multi-LLM evaluation across Claude, Gemini, and OpenAI.
Designed and operated internal services for model and provider access with strong access controls, secret management, audit logging, and automated credential rotation — enabling self-service use without losing governance.
Migrated the Azure OpenAI Services proxy from Apigee Edge to Apigee X, improving security posture, scalability, and operational monitoring for downstream AI workloads.
Developed reusable Terraform and Ansible modules and CI/CD patterns for internal platform programs that standardize application and AI service delivery, making platform work repeatable, reviewable, and easier to operate over time.
Built self-service GPU and HPC environments using SLURM and GCP Managed Instance Groups, including MIG-based partitioning and distributed deployments via Google Cluster Toolkit to support scalable training and data processing workloads.
Integrated GPU telemetry and health signals (NVIDIA DCGM + Cloud Operations Suite) to support SLO-driven operations, faster debugging, and better capacity planning for AI and ML workloads.
Built audit logging, reproducible pipelines, and HIPAA/PHI-aware workflows into AI services from day one — partnering with research, clinical, finance, and cloud operations teams to ensure designs meet security, compliance, and real-world usage requirements.
Strong defaults, automation, and observability reduce friction without sacrificing reliability or trust. When platforms provide clear guardrails and visibility, engineers spend less time fighting infrastructure and more time shipping.
That approach scales — in big-tech environments and regulated domains — because it is fundamentally about operational clarity.
The core technologies and systems I work with across cloud platforms, AI services, distributed compute, and ML infrastructure.
Mayo Clinic · Generative Artificial Intelligence Program
Remote · 2024 to PresentI design and build cloud-native platform services on GCP for internal platform programs that standardize application and AI service delivery, developing reusable Terraform and Ansible modules and CI/CD patterns. I build secure backend services and admin UIs for managing model access and credentials, migrated an Azure OpenAI proxy from Apigee Edge to Apigee X, and develop GPU/HPC self-service capabilities using SLURM and Managed Instance Groups, paired with a GPU observability stack built on Cloud Operations and NVIDIA DCGM. I also contributed core components to a Patient MCP Server, a FastAPI-based Model Context Protocol service exposing clinical data (FHIR, BigQuery) to LLM-driven agents, with tool routing, LOINC/ICD code integration, and multi-LLM evaluation across Claude, Gemini, and OpenAI.
Mayo Clinic · Multimodal Neuroimaging Laboratory
On-site · 2021 to 2024I built high-throughput pipelines for diffusion MRI and intracranial EEG processing using Python, Linux, and distributed workflows. I led development of HED-SCORE, an open-source EEG metadata framework adopted by international research teams, and designed tools for multimodal data integration, structured metadata management, and quality control across large neuroimaging datasets in collaboration with clinicians and data scientists.
Mayo Clinic · Bioelectronics Neurophysiology and Engineering Laboratory
On-site · 2018 to 2021I worked on ML for long-duration EEG and wearable sensor data, developing LSTM-based seizure prediction models, integrating real-time ML components into clinical-grade monitoring systems, and building ingestion, preprocessing, and feature extraction pipelines for large biosignal datasets in collaboration with neurology and engineering teams.
Mayo Clinic · Bioelectronics Neurophysiology and Engineering Laboratory
On-site · 2016 to 2018I prototyped closed-loop neurostimulation systems that combined hardware signals with real-time software control, and engineered backend tools for structured patient data tracking, automation, and biosignal analysis while supporting research teams with software development, data workflows, and database design for experimental studies.
Model Context Protocol service exposing clinical data to LLM-driven agents. Includes tool routing across backends, LOINC and ICD code integration, and multi-LLM evaluation across Claude, Gemini, and OpenAI.
Internal service for managing OpenAI service accounts and API keys. Integrates with GCP Secret Manager, includes audit logging and scheduled rotation, and provides a simple admin UI for platform and security teams.
Migrated the Azure OpenAI Services proxy from Apigee Edge to Apigee X, improving security posture, scalability, and operational monitoring for downstream AI workloads.
Unified GPU telemetry stack that collects metrics and health signals from GPU nodes and exposes them in dashboards and alerts. Enables teams to track utilization, debug issues, and plan capacity for AI and ML workloads.
Automated system for creating and managing SLURM clusters on demand for research and production teams, using instance templates, Google Cluster Toolkit, and infrastructure automation to standardize configuration and simplify lifecycle operations.
Reusable infrastructure modules and CI/CD pipelines supporting consistent deployment of cloud-native applications and AI-enabled services across internal platform programs.
Machine-readable metadata framework for EEG annotations that supports FAIR-aligned data sharing and large-scale analysis. Designed to make neurophysiology datasets easier to reuse across labs and tools.
View on Google ScholarPython-based pipelines for diffusion MRI and intracranial EEG preprocessing and analysis. Built to support reproducible studies and collaboration between engineering and clinical research teams.
Models and algorithms for long-duration EEG and wearable sensor data, including LSTM-based seizure prediction, anomaly detection, forecasting of biomedical time series, and signal artifact reduction.
View on Google ScholarSystem for seizure risk forecasting using EEG and wearable data. Combines feature engineering, temporal modeling, and long-horizon prediction for real-world monitoring scenarios.
View on Google ScholarConnect with me to discuss roles or collaborations related to cloud platforms, AI services, MCP and agentic systems, and ML infrastructure.
Open LinkedIn ProfileBrowse my publications and patent work related to seizure prediction, EEG metadata, and multimodal neuroimaging.
View Google ScholarIf you’re building cloud platforms, AI services, or secure AI infrastructure — I’d love to connect.
Email Me