Builder · Founder · Operator · Architect
20+ years across public and private, corporate America, venture capital, and small businesses. Enterprise technology director, leader of engineering organizations, hands-on platform/product architect, and a managing founder of my own ventures.
Three principles that have shaped every platform I’ve built and every team I’ve led.
I’ve built companies, products, and teams from zero — and I’ve never stopped. Continuous building is my founder reflex: I own the outcome, ship something real, then iterate.
Architecture earns its keep when the product serves the customer. I design from the outside in, backwards into strategy. Choices follow the product, not the other way around.
Twenty years × eight domains × four roles — builder, founder, operator, and architect. I recognize a problem’s underlying pattern, then optimize the solution.
Selected large-scale initiatives where I led architecture, design, and delivery across complex enterprise environments.
Built a governed, cloud-native integration platform that gives product teams a single, consistent way to exchange data across the portfolio — supporting both event-driven and request/response patterns, with built-in routing, durability, and isolation across products, tenants, and environments. Built to scale with the business without each product team having to solve integration from scratch.
Built a 50+ person full-stack technology organization from zero — Help Desk through Engineering — with P&L ownership of $10M+ across 125+ sites and 30+ customers. While staffing the org, we shipped two foundational builds in parallel: a full industry platform with native mobile and web front-ends, and the merger of two physical datacenters into a hybrid cloud. The result: insourced previously outsourced services, 15% off operating budget, and product uptime from 80% to 99%.
Designed and implemented enterprise data platforms across AWS and Azure tech stacks — building transformation pipelines that move data from transactional systems through normalized operational layers into analytical/BI presentation, agentic RAG retrieval, and predictive analytics, all drawing from a single source-of-truth with a reusable foundation the business could grow into.
Technical architect for one divestment — designing how to separate two companies sharing the same identity systems and migrate the divesting entity onto an independent platform stack, with clean cutover of people, systems, and capabilities. Participated in an acquisition and a public offering marketing roadshow. On the acquisition side, helped integrate workers, workflows, and systems into the existing operating environment.
Led the consolidation of two physical datacenters into a hybrid cloud environment serving 100+ edge locations with local-first HADR.
Led the full decomposition and replatforming of a legacy monolithic order management system into a scalable, cloud-native managed SaaS platform using serverless and containerized architecture.
Designed and delivered a centralized multi-product, multi-IdP identity broker supporting federated SSO, OIDC, and OAuth2 with both online and offline capabilities across multiple environments.
Led development of an ML + AR-powered mobile label scanning system for warehouse and logistics operations.
Designed and delivered enterprise DevOps CI/CD pipelines for Engineering and Infrastructure teams, replacing legacy Jenkins with a modern SaaS-based solution.
Served as lead architect and developer for an internal enterprise BPA platform spanning telematics, mobility, safety, and regulatory compliance — replacing manual workflows.
Migrated a legacy EDI (X12) platform to a cloud PaaS-based API solution, modernizing B2B data exchange across 30+ customers.
Built enterprise BI dashboards, KPIs, and reporting across operational and data warehouse environments, plus a full data catalogue.
Led development of ML predictive models for operational forecasting and decision-making within a transportation environment.
Drove a UI rejuvenation strategy using reusable Figma components to accelerate low-code/no-code frontend development across the organization.
Architecting AI into products is a hands-on craft. The lab is where patterns get debugged before they get recommended.
vLLM + Ray cluster · 200GbE RoCE interconnect
A pair of NVIDIA DGX Sparks in my home office is how I prototype, experience, and strategize before recommending anything to anyone. Real models, real workloads, real failure — the kind of fluency you can’t fake.
This page is updated automatically by a scoped agentic context-gathering process running on the lab below — an agent harvests signal from my repos, projects, and notes, scopes it to the slug of this portfolio, and submits proposed edits. The portfolio you’re reading is itself an output of the architecture it describes.
A custom gateway sits in front of all requests — direct or agentic — classifying complexity, then dispatching with the right model config (thinking vs. fast, temperature, context budget) for the job.
Weight precision, KV-cache compression, MoE expert pruning (REAP), and serving throughput — preserving model quality while local dense models get small and fast enough to serve in production.
Grounding agents in source-of-truth data — every retrieval scoped to a specific tenant or customer slug so cross-tenant leakage never happens. The same source-of-truth pattern the Enterprise Data Platform applies at scale.
Qwen3.5-122B is the sweet spot for this hardware, REAP-pruned for serving throughput without the quality cliff. The custom gateway router decides per request when a heavier reasoning pass earns its tokens.
Hermes is the harness in play today, deployed behind a swappable adapter so the harness layer can change without touching anything upstream. Each instance is scoped to a specific slug — tenant, customer, or audience — so a harness serves one defined population, not the whole house.
The public protocols each solve one slice of agent communication. Agency Protocol (AIDP-C) adds the missing piece: cryptographically verifiable agent identity and scoped delegation. Open standard, designed to work alongside ACP, UCP, A2A, and FIDO agentic flows.
Self-hosted frontier models, agents, and retrieval at production scale
Multi-cloud platform builder — fluent across both hyperscalers, comfortable in hybrid, on-prem, and edge
Turning raw data into actionable intelligence at scale
Hands-on across the polyglot stack — transactional, analytical, and embedded
Standardized build, test, and ship for engineering and infra alike
Building trust, federation, and compliance into every layer — for humans and agents alike
A foundation in business and science, sharpened by industry certifications across cloud architecture and security.