Principal Architect, ChenMed LLC

Architected and built a fully automated AI-driven patient outreach system that calls and texts high-risk Medicare Advantage patients, scores every conversation in real time for clinical risk, and hands off patients in need to live clinical staff for immediate follow-up. Designed a novel dialog engine on top of third-party telephony and speech services, operating within HIPAA and clinical-safety constraints. Built the 5,000-response vector store from prior call transcripts and clinical input — a proprietary NLP asset capturing how this patient population communicates about their health. First live call within one week of project start. Now runs at ~4,000 patient conversations per day across voice and SMS, with 500+ daily escalations to clinical staff at 68% audited precision — ~125 ER visits prevented and ~$2M in estimated medical cost reduction to date.

Case study →

Principal Architect, ChenMed LLC

Built a production system that ingests clinical documents from specialists, labs, ERs, and fax — over one million files per week across XML, PDF, and scanned image formats — and identifies new or changed patient conditions using a layered approach: vector embeddings for bulk triage, Claude Sonnet for low-confidence documents that require deeper reasoning. Automatically adds high-confidence findings directly to patient charts; routes others to clinical review. In a value-based care model, accurate and timely condition capture is foundational to both patient safety and risk-adjusted capitation revenue. Designed to scale: rebuilt ingest layer from file-polling to Snowpipe + SQS event notifications when volume reached 7× the original design target.

Case study →

Principal Architect, ChenMed LLC

Designed and built a near-real-time CDC engine replicating the clinical system of record into Snowflake — 1.88 billion rows across 140 tables, 3.5 million change events per day, 15-second ingest latency, 99.997% task reliability. Replaced commercial CDC tooling with an open-source-based pipeline (MySQL binlog → bespoke Python CDC service → Snowflake Streaming API → serverless task DAG → tabular view layer) that handles the merge semantics, delete fidelity, and schema evolution problems that vendor tools solve expensively and imperfectly. Data is shared live to four downstream Snowflake accounts at approximately $65 per million events applied, with near-zero marginal cost per additional table. A nightly File Factory backstop guarantees convergence. The DBA team now uses the Snowflake tables as a warm disaster-recovery standby — organic adoption that confirmed data quality confidence.

Case study →

Senior Support Engineer, Amazon.com

Authored the internal migration runbook for Amazon's MAWS initiative — the company-wide mandate to move applications from VMs onto AWS infrastructure. Provided the prescriptive mapping of modules and subsystems to the available AWS primitives of the era (EC2, S3, EBS, RDS) and served as the designated SME for VM-to-AWS migrations across the Merchants@ organization. The runbook became the foundation for subsequent migrations of both internal and customer-facing systems. This work predates the managed-service ecosystem most AWS engineers now take for granted; the migrations were built from cloud primitives, which required architectural judgment that later abstractions made optional.

Case study →

Senior Principal Consultant, Oracle Corporation

Architected and built a server orchestration abstraction layer managing the full lifecycle — provisioning, patching, configuration, disaster recovery, decommissioning — across 110,000 servers, 20 data centers, and seven regions spanning public cloud, FedRAMP, and private cloud. Designed the APIs, CI/CD pipelines, and SaltStack-integrated automation that enforced security and compliance at federal scale. Co-invented U.S. Patent #15/681,084 for electronic voting security. Three-time recipient of the Oracle Fusion Ops Engineering Excellence Award (2015, 2016, 2017).

Case study →

Senior Database Architect, Everyone Counts

Led the migration of New Jersey's voter registration system from 24 Oracle 9/10/11 instances to MariaDB 10.1 under a $12M contract, delivered three months ahead of schedule. Designed a 6TB, 13-site, multi-directional, fault-tolerant multi-master replication topology allowing simultaneous writes across all counties while surviving frequent network partitions. Built custom Perl ETL to migrate Oracle data into the new architecture, converted the platform from Windows to Linux, and trained county SysAdmins and DBAs on the new stack. The platform has since maintained 100% uptime through hundreds of local and state election cycles.

Case study →

Lead Developer and Founding Team Member, Netmark Systems

Architected and led development of a full multi-tenant platform serving the direct-sales industry — CRM, replicated distributor websites, email marketing automation, compensation-plan management, payment processing, ecommerce, and online training systems. Operated as white-label backend for entire customer organizations, with individual customers running dozens of distinct brands. Scaled past 600,000 lines of Perl on 25+ Linux servers, built the marketing automation engine that delivered 22M+ messages daily, and helped grow the company to 50+ employees and $12M+ annual revenue in three years — predating the vertical-SaaS category by nearly a decade.

Case study →

Led data engineering and production AI architecture for a national Medicare Advantage managed care operator serving 250,000+ active patients across 100+ primary care centers. Responsible for the full data platform: real-time clinical data replication, high-volume document ingestion, production AI systems, and the data infrastructure supporting analytics, clinical operations, and executive reporting. Built File Factory — a configuration-driven data movement platform (Celery + Redis, Snowflake) that has processed 12 million files, 439 billion database records, and sustains 7.4 billion rows of movement per day across 10 source systems and 700+ destination objects. See selected engagements above for the AI and CDC work.

Primary technical advisor for MariaDB's largest Fortune 100 and government accounts. Drove $13M in ARR over 19 months. Engaged with 65TB+ deployments running at 100,000+ TPS, providing architectural guidance on ColumnStore, Galera Cluster, and MaxScale for high-throughput financial, government, and healthcare workloads. Designed and reviewed migration strategies from Oracle, MySQL, and commercial databases to MariaDB for enterprise clients. Represented MariaDB in pre-sales engagements, technical evaluations, and production incident escalations for the most demanding accounts in the portfolio. MariaDB 10.3 Certified DBA.

See selected engagement above: Global Server Orchestration for Oracle Fusion SaaS. 110,000 servers, 20 data centers, 7 regions, FedRAMP. Co-inventor, U.S. Patent #15/681,084. Three-time Engineering Excellence Award.

See selected engagement above: Statewide Election System Migration, State of New Jersey. $12M contract, delivered three months ahead of schedule. 13-site multi-master MariaDB replication topology; 100% uptime through hundreds of election cycles.

Built the MySQL automation framework — Chef-based infrastructure-as-code for configuration management, dynamic hardware-aware tuning, replication topology management, automated host rebuilds, and backup recovery — that became the foundation of HP's commercial Database-as-a-Service public cloud offering. Managed 30+ production MySQL instances running at 6,000+ concurrent connections and 10,000+ QPS. Closed SOC compliance gaps in access control, modification logging, and backup recovery verification. Designed explicitly for handoff: the automation was complete and self-documenting enough that the successor described it as the easiest MySQL environment they had ever inherited.

See selected engagement above: Amazon MAWS migration runbook authorship. SME for VM-to-AWS migrations across the Merchants@ organization.

Architected and operated the database infrastructure behind a major options trading platform — multi-master, heavily sharded MySQL across two production data centers, 2TB+ clusters sustaining 7,000+ QPS, a 1-second order-processing SLA, and 100% uptime during trading hours. Designed the replication strategy for trading accounts, financial instruments, and live position data in the pre-synchronous-replication era. Built database monitoring and SMS alerting independent of data-center connectivity. Evaluated and ultimately rejected a commercial synchronous-write proxy after it failed under production load. Architected the platform as Tavant Technologies' Director of Architecture (2007–2009), then was hired directly by OptionMonster to operate it.

Live migration of millions of mailboxes from mbox to Maildir storage without mail loss, while replacing the webmail product. Built an 11TB, 90-day rolling MySQL HTTP-log warehouse before columnar warehouses were viable — MERGE-table partitioning, MEMORY-engine ingest staging, IOPS-aware hardware design. Moved a data center using bulk disk shipment plus distributed agents for delta synchronization.

See selected engagement above: Vertical SaaS Platform for Direct Sales Organizations. 600,000+ lines of Perl, 22M+ messages/day, 1,000+ domains served from a single Apache instance, $12M+ annual revenue.

Real-time financial transaction systems for an online gaming operator: payment processing, identity verification, and fraud detection. Built the replicated branded-site architecture later scaled at Netmark.