ChenMed LLC

Production AI products for a national Medicare Advantage operator serving 250,000+ patients across 100+ primary care centers.

• Shipped an AI patient-outreach system in one week. Designed and built Grace, a fully automated system that calls and texts high-risk Medicare Advantage patients to check on their wellbeing. The first live call went out within a week of project start: no vendor procurement, no pilot committee.
• Scaled it to ~4,000 patient conversations per day. Launched at 300 calls per hour with 5,000 scored conversations in the first week; now runs in sustained production at roughly 4,000 conversations a day.
• Prevented ~125 ER visits and ~$2M in medical cost. Measurable reductions in hospital admission rates appeared within days of launch. To date the system has averted an estimated ~125 emergency-room visits and ~$2M in medical cost.
• Built real-time clinical-risk scoring with a human handoff. Every conversation is scored as it happens; patients in need are escalated to live clinical staff for immediate follow-up — 500+ escalations per day, audited at 68% precision.
• Designed a novel dialog engine on commodity telephony. Built the conversation engine on top of third-party telephony and speech services, operating inside HIPAA and clinical-safety constraints rather than adopting a turnkey voice-bot product.
• Extended the system from voice to SMS. Added text messaging as a patient-selected alternate channel — asynchronous threaded conversations, longer time-to-resolution windows, and different pacing, all on the same dialog engine and risk scoring.
• Turned prior call transcripts into a proprietary NLP asset. Built a 5,000-response vector store from historic call transcripts and clinical input, capturing how this patient population actually talks about its health.
• Targeted outreach by clinical risk, not call volume. A daily warehouse query selects the highest-risk patients at each care center, ranked by CMS-governed HCC risk scores tied to verified clinical conditions.

Stack Agentic dialog systems · vector embeddings · real-time conversational scoring · telephony & speech APIs · Snowflake · HIPAA

ChenMed LLC

Organization-wide data platform transformation, and the first AI products built on top of it.

• Consolidated fragmented legacy architectures onto Snowflake. Drove organization-wide adoption of a centralized, governed platform: 1.6PB of data, 1,000+ users, and 3,000+ DAG steps across seven Snowflake accounts.
• Built File Factory, a configuration-driven data movement platform. Celery + Redis on Snowflake. Has processed 12 million files and 439 billion database records, and sustains 7.4 billion rows of movement per day across 10 source systems and 700+ destination objects.
• Replaced commercial CDC tooling with a purpose-built engine. Near-real-time change data capture from 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. Handles ordering, idempotency, delete fidelity, and schema evolution across five heterogeneous source systems.
• Brought CDC cost to ~$65 per million events. MySQL binlog → bespoke Python CDC service → Snowflake Streaming API → serverless task DAG, with near-zero marginal cost per added table. Shared live to four downstream Snowflake accounts; a nightly backstop guarantees convergence.
• Earned adoption as the warm disaster-recovery standby. The DBA team now relies on the Snowflake replica as a DR standby: organic adoption that confirmed confidence in the data’s fidelity.
• Launched clinical document intelligence over 1M+ files per week. Built High-Risk Disease Detection: ingests ER records, specialist notes, labs, faxes, and third-party documents across XML, PDF, and scanned TIF, and matches them against ICD-10, HCC, and high-risk condition taxonomies — cancers, diabetes, cardiovascular and renal disease.
• Surfaced 12,000+ clinical recommendations in two weeks. Findings that would otherwise have been missed or delayed. Layered design: vector embeddings for bulk triage, Claude Sonnet for low-confidence documents; high-confidence findings post to patient charts, the rest route to clinical review.
• Re-architected ingest when volume hit 7× design. Rebuilt the document ingest layer from file polling to Snowpipe + SQS event notifications without interrupting the clinical workflow.
• Unified ingestion for every inbound clinical format. Designed the ingestion layer for HL7, FHIR, EDI, Kafka, XML, JSON, Excel, PDF, and API sources.

Stack Snowflake (multi-account, Cortex, Snowpipe, Streaming API) · Python · Celery/Redis · AWS SQS · vector embeddings · Claude · HL7/FHIR/EDI

ChenMed LLC

Hired to overhaul a MySQL-centric IT operation running the in-house EMR and modernize its data infrastructure.

• Upgraded the clinical EMR database from MySQL 5.7 to 8.0. Planned and executed the major-version upgrade of the system of record for clinical operations.
• Brought enterprise access control and audit to the database tier. Integrated Active Directory and role-based access control, and deployed MySQL Enterprise Audit for compliance visibility.
• Put schema changes under CI/CD. Implemented Jenkins- and Liquibase-based pipelines so database changes are versioned, reviewed, and repeatable.
• Simplified the AWS footprint. Optimized the AWS architecture to remove complexity and improve enterprise-wide data reliability.
• Built the MySQL-to-Snowflake replication layer in-house. A near-real-time ELT pipeline that enabled the move to cloud data warehousing at a fraction of the cost of commercial alternatives, and became the foundation for the org-wide Snowflake adoption that followed.

Stack MySQL 8 · MySQL Enterprise Audit · Active Directory/RBAC · Jenkins · Liquibase · AWS · Snowflake

MariaDB Corporation

Primary technical advisor for MariaDB’s premier Fortune 100, media, healthcare, defense, and government accounts.

• Drove $13M in ARR in 19 months. Crossed regularly from architecture into sales engineering — pre-sales pitches, technical evaluations, and proofs of concept for the most demanding accounts in the portfolio.
• Advised a marquee client roster. Primary or secondary technical accountability for accounts including Disney, NBC Universal, Walgreens, CVS, Latham & Watkins, TiVo, Bank of Canada, Fidelity, Quotient Technologies, ICANN, Lockheed Martin, and BAE Systems.
• Carried a concurrent book of five premium TAM accounts. Each a $150K/year technical account management relationship, sustained through 50% travel to client sites.
• Architected 65TB+ deployments at 100,000+ TPS. Architecture review and operational guidance for high-throughput financial, government, and healthcare workloads on ColumnStore, Galera Cluster, and MaxScale.
• Designed migrations off Oracle and commercial databases. Designed and reviewed migration strategies from Oracle, MySQL, and other commercial engines to MariaDB for enterprise clients.
• Owned the hardest production escalations. Represented MariaDB in production incident escalations for its most critical accounts, covering encryption, multi-threaded and multi-master replication, Galera synchronized commit, auditing, and performance.
• Taught and spoke for the company. Delivered on-site and remote customer training, spoke at MariaDB’s NYC conference, and built automation and reference-architecture repositories used across engagements.

Stack MariaDB · ColumnStore · Galera · MaxScale · Oracle/MySQL migration · MariaDB 10.3 Certified DBA (2019)

Oracle Corporation

Fusion SaaS operations during Oracle Fusion’s transition from shipped software to a cloud service.

• Put change control around 60,000 hand-run scripts. Inherited an estate of version-specific install, patch, backup, and failover scripts that operators ran from runbooks by copy-paste. Built a human-gated CI/CD pipeline so every change was reviewed, auditable, and reversible before anything was replaced.
• Retired the script estate with an orchestration abstraction layer. Unified provisioning, configuration, hydration, patching, disaster recovery, and decommissioning into a single lifecycle, eliminating 60,000+ legacy scripts and reducing operational overhead for 1,000+ offshore operations engineers.
• Scaled orchestration to 110,000 servers. Across 20 data centers and seven regions spanning public cloud, FedRAMP, and private cloud.
• Enforced security and compliance at federal scale. Designed the APIs, CI/CD pipelines, and SaltStack-integrated automation that kept FedRAMP environments compliant by construction.
• Attacked the labor cost eating Fusion Cloud’s margins. Automation replaced manual hardware allocation and scripted setup for every new customer, the largest operational cost in the shift from on-prem installs to SaaS.
• Co-invented U.S. Patent #15/681,084. Electronic voting security (Oracle-owned).
• Won Engineering Excellence three years running. Oracle Fusion Ops Engineering Excellence Award, 2015, 2016, and 2017. Guest lecturer for UC Riverside’s MIS 411 Database Management Systems (March 2018).

Stack SaltStack · CI/CD for operations · Python · FedRAMP · public/private cloud · Oracle Fusion Apps

Everyone Counts

Election technology. Database lead on the State of New Jersey’s statewide voter registration rebuild.

• Delivered a $12M state contract three months early. Migrated New Jersey’s voter registration system, covering 5M+ registered voters, from 24 Oracle 9/10/11 instances to a ground-up rebuild on distributed MariaDB 10.1.
• Designed for writes during network partitions. Built a 6TB, 13-site, multi-directional multi-master replication topology so every county could keep accepting writes through the frequent network outages of election infrastructure.
• 100% uptime through hundreds of election cycles. The platform has run without an outage through every local and state election since launch.
• Migrated the data and the platform. Wrote custom Perl ETL for the Oracle-to-MariaDB conversion and moved the platform from Windows to Linux, paired with a Java and C#/.NET application tier.
• Led the database team and trained the operators. Led three junior DBAs alongside the software architect, and trained county SysAdmins and DBAs who would run the system on election night. Made multiple on-site visits to county election offices.
• Automated provisioning and testing. Chef cookbooks for all database provisioning with source-controlled schema versioning, plus an automated performance-testing environment for every client system.
• Designed platforms for new jurisdictions. Sales engineer and designer for voter registration platforms pitched to New Jersey, Florida, Michigan, Hawaii, Canadian provinces, and Australia, plus private-sector voting for the Emmys and the Academy Awards.
• Integrated with state agencies. Designed data-sharing integrations with DMV, Social Security Administration, and Prison & Parole systems.

Stack MariaDB 10.1 multi-master · Oracle 9/10/11 · Perl ETL · Chef · Linux

HP Cloud Services

Database Infrastructure Service team running the MySQL tier behind HP Public and Private Cloud.

• Built the automation that became HP’s commercial DBaaS. A Chef-based MySQL framework (configuration management, hardware-aware tuning, master-slave and multi-master replication, host rebuilds from backup images, scheduled backups) that became the foundation of HP’s public-cloud Database-as-a-Service offering.
• Ran MySQL at cloud-provider scale. 30+ production and 50+ dev/test instances totaling 13TB, some sustaining 6,000 concurrent connections and 10,000+ QPS, across a 10,000+ host estate in five data centers spanning AMER, EMEA, and APAC.
• Kept it running 24×7. Established and carried the round-the-clock on-call rotation for the production service with one other engineer.
• Closed SOC compliance gaps. Identified and remediated technical and procedural gaps in access control, modification logging, and backup-recovery verification.
• Won enterprise deals against Oracle. Built proof-of-concept environments and proposal guidance for HP Enterprise Sales, showing horizontally scaled MySQL replacing Oracle for datasets growing 1TB/month at significant licensing savings.
• Automated myself out of the job. Designed for handoff from the start. The successor called it the easiest MySQL environment, and the easiest job, he had ever inherited.

Stack MySQL · Chef · multi-master replication · SOC compliance · public & private cloud

Amazon.com

Merchants@ organization, during Amazon’s company-wide mandate to move internal applications onto AWS.

• Wrote Amazon’s internal runbook for the move to AWS. Authored the MAWS migration runbook: a prescriptive mapping of application modules and subsystems to EC2, S3, EBS, and early RDS, written before CloudFormation, Lambda, or most managed services existed. It became the foundation for subsequent internal and customer-facing migrations.
• Served as the org’s VM-to-AWS migration SME. Designated subject-matter expert for migrations across the Merchants@ organization.
• Cut ticket volume 20% in two quarters. Built deduplication and suppression trees into the alerting infrastructure, with automated alert suppression during high-severity incident response.
• Built recruiting analytics and supported SVP-level reviews. Delivered a recruiting analytics framework and operational metrics for SVP-level operational reviews.
• Earned four awards in nine months. M@ Operational Excellence Practitioner, Anvil Security Award, MAWS Program Guide, and Distinguished Engineer Award. Hired at L5, the top of band, with an offer within hours of a VP-level interview close.

Stack AWS (EC2, S3, EBS, RDS) · incident response · alerting · operational metrics

OptionMonster Holdings

Hired directly by the client to operate the options-trading database platform I had architected as a Tavant contractor.

• Was hired by the client to run what I built. After two years as Tavant’s Director of Architecture on the engagement, OptionMonster hired me directly to operate the platform.
• Met a 1-second order-processing SLA with 100% trading-hours uptime. Operated 2TB+ multi-master, heavily sharded clusters sustaining 7,000+ QPS across two production data centers, with multi-data-center disaster recovery.
• Built a real-time NYSE tick database. Captured market ticks for predictive and analytical applications.
• Built monitoring that survived data-center outages. Co-built a fully in-house monitoring, reporting, alarming, and self-healing system in Python, with SMS alerts sent from a cellular modem so they went out even when data-center connectivity didn’t.
• Saved six figures a year by being the support contract. Ran the platform without a commercial Sun MySQL Enterprise support agreement; internals expertise covered what the vendor contract would have.
• Automated database deployments before "CI/CD" was common. Subversion-driven schema deployment with execution-order files and manual gating for production.
• Covered Oracle too. Sole Oracle 10g/11g DBA for nightly batch operations: installation, administration, and maintenance.

Stack MySQL multi-master & sharding · Oracle 10g/11g · Python · Subversion

Tavant Technologies

Led database architecture for Tavant’s build of the optionMONSTER.com options-trading platform under an open-source mandate.

• Architected a trading platform’s data tier from scratch. Multi-master, heavily sharded MySQL for trading accounts, financial instruments, and live positions across two Chicago production data centers and a Bangalore development site.
• Kept writes safe before synchronous replication existed. Designed the replication strategy for financial data in the pre-synchronous era. Evaluated a commercial synchronous-write proxy, spent months in Bangalore debugging it under load, and had it rejected and refunded when it failed.
• Tuned around ORM-generated SQL. Extensive query tuning on a Java/Hibernate stack; diagnosed and solved critical performance issues in core trading transactions and reporting.
• Built the team’s MySQL depth. Mentored Tavant’s software engineers and junior DBAs on MySQL internals, and trained a Bangalore engineer from scratch into the platform’s DBA.
• Sales-engineered for new trading clients. Pitched low-latency, high-throughput MySQL architectures to prospective Tavant clients in securities trading. Also supported EA’s The Sims Online community website and EA Download Manager.

Stack MySQL multi-master & sharding · Java/Hibernate · Linux

United Online (Freeservers)

Consumer web hosting and email at scale.

• Migrated millions of mailboxes without losing mail. Live migration from mbox to Maildir storage while simultaneously replacing the webmail product.
• Built an 11TB warehouse before columnar warehouses existed. A 90-day rolling MySQL 5 HTTP-log warehouse: MERGE tables standing in for native partitioning, MEMORY-engine ingest staging, MyISAM durable storage, and IOPS-aware hardware design.
• Moved a data center with autonomous agents. Bulk data shipped by truck on disk, with distributed agents synchronizing deltas until cutover.

Stack MySQL 5 · object-oriented Perl · Linux · mail storage

Netmark Systems

Multi-tenant, white-label platform for direct-sales organizations, a decade before "vertical SaaS" was a category.

• Helped grow the company to $12M+ revenue and 50+ employees in three years. Architected and led development of the platform the business ran on, serving dozens of direct-sales clients ranging from hundreds to hundreds of thousands of sales reps, some running dozens of brands each.
• Built a full vertical platform end to end. CRM, replicated branded distributor websites, compensation-plan management, payment processing, retail and business-opportunity ecommerce, online training, lead distribution, and in-house video conferencing.
• Delivered 22M+ marketing messages per day. Built the email marketing automation engine, with deliverability optimization, A/B testing, and engagement tracking, on early-2000s hardware.
• Served 1,000+ domains from a single Apache instance. Wrote a custom mod_perl handler that replaced Apache’s virtual-host resolution, sidestepping the file-descriptor exhaustion that capped stock configurations.
• Built a template engine ahead of its time. Nested SSI-style templates with RPC and SOAP content fetching, structurally close to what later became edge-side includes.
• Scaled the codebase and the fleet. 600,000+ lines of Perl on 25+ Linux servers, with at least one client system tracking 10M+ leads.

Stack Perl · mod_perl · Apache · MySQL · Linux · SOAP

Casino Ventures

Online gaming and sportsbook operator.

• Built real-time financial transaction systems. Payment processing, identity verification, and fraud detection for a high-volume, heavily regulated money flow.
• Originated the replicated-site architecture. Built replicated branded affiliate sites, the pattern later scaled to 1,000+ domains per server at Netmark.

Stack Perl · payments · fraud detection

Charter Communications

Fortune 100 cable, broadband, and mobile operator.

• Automated lower-environment databases for ServiceNow. Wrote Python tooling to manage the dev, test, staging, and QA databases behind Charter’s internal ServiceNow instance.
• Handed it off clean. Documented the tooling and transitioned it to the team that inherited it.

Stack Python · ServiceNow · database automation

CellTrak Technologies

Mobile workforce platform for home-health and hospice agencies, under HIPAA. Brought in for operational rescue.

• Won the engagement with a pitch to management. Prepared and delivered the proposal that secured the work directly.
• Fixed serious performance problems on very large RDS MySQL. Tuned high-velocity mobile write ingestion and read-heavy reporting within RDS instance, IOPS, and storage limits.
• Led recovery from a complete production outage. On site during the outage; restored service working alongside AWS engineering teams.

Stack AWS RDS · MySQL · HIPAA

Serenova

Short contract on MySQL replication.

• Supported MySQL replication with Tungsten Replicator. Short-term replication work for the contact-center software company.

Stack MySQL · Tungsten Replicator

GCI Communication Corp

Alaska’s largest telecommunications and cable provider.

• Automated GCI’s MySQL footprint with Chef. Most of the engagement went to infrastructure-as-code for provisioning and managing MySQL at telecom scale, alongside ongoing database management.

Stack Chef · MySQL

Naxum LLC

Rebuild work on a multi-tenant direct-sales platform that now serves 1M+ users.

• Stood up a development team in Bucharest, Romania. Traveled to Bucharest, navigated the local legal environment, leased office space, hired local engineers, and managed the team remotely to make delivery economical.
• Built a private-cloud HA platform. Migrated from three multi-duty servers to a fully redundant Xen environment: 36 VMs across three hosts.
• Built a new CMS platform. MongoDB, jQuery, and Bootstrap over a Perl service architecture that separated business and presentation logic.
• Put a mobile front end on 13-year-old code. A JSON service layer and jQuery Mobile site that became the primary mobile "virtual office," with full backward compatibility.

Stack Perl · MongoDB · Xen · jQuery Mobile

Multiply.com

Social-commerce platform pivoting to ecommerce in Southeast Asia after its acquisition by Naspers.

• Re-engineered 150–250+ shards for availability. Brought MySQL shards of 4–80GB each, never designed for HA or fault tolerance, up to industry standard.
• Achieved PCI-DSS readiness for ecommerce. Re-engineered systems to meet PCI-DSS as the business moved into payments.
• Removed single points of failure. Replaced stack-based workloads with RabbitMQ message queues throughout the environment.
• Rebuilt monitoring and reporting. Async metrics collection from 250+ MySQL instances into a Cassandra cluster (250–300 writes/sec), and business reports moved from transactional queries to materialized views and ETL.
• Wrote the handover manual. Authored the SOP manual for the operations transfer to the new Indonesian headquarters.

Stack MySQL sharding · Cassandra · RabbitMQ · Perl · node.js · PCI-DSS

MightyPages.com

Hyper-local shopping portal. Lead architect from inception.

• Architected the platform from inception. Object-oriented Perl n-tier design, with backend services consumed by a PHP front end. The platform remains in operation.
• Modeled any product or service on earth. A relational data model general enough to describe any product or service, proven with ~2M products, with heavy geospatial search over a highly abstracted schema.
• Ran it highly available. Linux HA on Xen VMs over MySQL and MongoDB.

Stack Perl · node.js · MySQL · MongoDB · Xen · geospatial search

City of Chicago

Public-health infrastructure.

• Built the database tier for infectious-disease tracking. MySQL clusters for a system that let local hospitals communicate securely when notable infectious diseases were detected, supporting transmission tracking and early identification of hot zones.

Stack MySQL clustering · public health

Dial-up ISP, Junction City, KS

Regional internet provider.

• Built web properties and light software. Web work for ISP customers, plus programming for a juvenile correctional facility.

KTUU (NBC Anchorage) and Absolute Internet

Concurrent roles at Alaska’s NBC affiliate and a local ISP.

• Ran the station website and grew its traffic. Maintained KTUU’s site and built traffic-generation tools, including Alaska e-cards, in Perl/CGI.
• Built sites for local businesses. Requirements to launch for Absolute Internet’s small-business clients.

ARCO Alaska

Hired at 16 by the comptroller’s organization to support the accounting floors; grew into building the business’s internal applications.

• Built an online flight-reservation system on Sabre. Integrated with the Sabre GDS around the time Travelocity launched, automating the corporate travel desk’s booking work.
• Built HR and travel intranet applications. Started in C CGI on Unix, then moved to Perl 5 and CGI.pm.
• Integrated mainframe HR data into the intranet. TN3270 screen-scraping integration, and an early, lasting lesson in handling sensitive data from Corporate Security.
• Instrumented the floors before users noticed. Self-initiated syslog server on surplus Apple PPC hardware running Linux, monitoring every floor printer so failures were caught before they were reported.
• Coordinated the Apple-to-PC corporate transition. Matched user capabilities with central IT and contractors during the migration.
• Built a GPS truck-tracking application for Schlumberger. Outside contract work: Visual C++, Access, and IIS CGI translating GPS coordinates onto a map of the Alberta fleet in real time. Also modeled a 4D database for an MSDS distribution business.