Back to overview
Professional experience

Two decades of building foundations

From the Education as a Service platform at Apollo, to real-time identity at Shutterfly, a consolidated AI and data platform at Pearson, and the enterprise AI strategy at Wood Mackenzie. In every role the job came down to building the foundation everything else runs on.

Two decades covers a lot of ground, and no two readers want the same slice of it. So this works the way a language model does when it loads a skill: pick a lens and the experience below re-renders against it, keeping only the work that matches and setting the rest aside. Start with everything, or narrow straight to the one thing you came to evaluate.

Depth

Click any role to expand or collapse it.

VP, Architecture, Platform Engineering and AI Wood Mackenzie San Jose, CA Mar 2025 to Present

Global energy analytics. A $500M product suite serving energy and utility companies.

  • Set the enterprise AI strategy from scratch across generative AI, data, classical ML and MLOps, and secured C-suite investment for a multi-year roadmap.
  • Set the architecture and led the team that built a proprietary agentic AI framework and platform on AWS AgentCore, Strands and Bedrock in under six months, using MCP and A2A for multi-agent orchestration. Agents retrieve trusted context, complete multi-step work and hand off to a person when judgment is needed. Live in the flagship Lens product and driving measurable sales growth.Building an Enterprise AI Platform14 minI wrote this up for AWSaws.amazon.com
  • Made the platform multi-model through a routing layer rather than betting the company on one vendor, sending each request to the model that fits the task, the cost ceiling and the latency budget, across OpenAI, Anthropic, Gemini and Bedrock, with failover when a provider degrades.
  • Developed a student-teacher distillation approach to produce a small language model, with Amazon Nova Pro as the teacher and Nova small as the student, trained on the digital content corpus, so inference cost and latency came down without surrendering answer quality.
  • Owned end-to-end AI lifecycle delivery: architecture, A/B testing and evaluation, deployment and production monitoring, tied to subscriber retention.
  • Built SageMaker MLOps pipelines with automated model CI/CD, a model registry and drift monitoring, so putting a model into production became a repeatable process rather than an event.Industrializing Forecasts17 min
  • Directed a Snowflake Medallion data architecture with Unstructured.io ingestion, vectorization and semantic re-ranking for retrieval, and built an AWS Neptune knowledge graph for North America power data.
  • Delivered the AI components the rest of the business builds on: LLM inference, similarity search and vector retrieval, guardrails, model evaluation, experimentation, governance and observability.
  • Integrated a Salesforce agent into the CRM workflow and exposed platform capability through GraphQL.
  • Built the data, AI, scraping and developer-enablement teams from the ground up, growing subject-matter experts already inside the company into the new domains and hiring specialists from outside where the gap was real rather than trainable, so the organization came to match the strategy instead of the strategy bending to the organization.One Runner, Many Patterns22 min
  • Pushed AI-accelerated development into the team's daily practice, using coding agents and assistants for scaffolding, migration and test generation under review, which is a large part of how a roadmap of this size was delivered by a team of this size in the time available.
  • Introduced a sprint operating rhythm that made delivery legible: scoped increments, demos every two weeks, and a reporting dashboard that aggregates Jira, Productboard, Confluence, GitHub and incident data into a single pane of glass, so status and blockers surface from the systems of record instead of being assembled by hand each month.The Architecture of Engagement18 min
  • Established AI governance standards through a Backstage developer portal and IAM and security architecture, and rolled out the Kiro IDE and Amazon Q so practitioners experienced the platform as acceleration rather than as control.
  • Ran AI literacy programs to embed responsible AI and a data-first culture across the business rather than only inside engineering.
  • Made the build-versus-buy calls across open source and commercial AI technology, and tuned production AI for cost, latency and throughput once it was carrying real traffic.
VP, Architecture, Platform Engineering and AI Pearson San Francisco, CA Jun 2020 to Mar 2025

A $750M product suite across MyLab, Mastering, Revel and MathXL. More than 250 engineers, globally distributed.

  • Led the convergence architecture for a $750M+ product suite and a globally distributed organization of more than 250, managing directors, managers, architects and senior technical leaders.
  • Owned hiring, retention, performance, coaching, succession and organizational design, and developed several engineers into managers, so the bench running delivery was one I had built rather than inherited.The Architecture of Engagement18 min
  • Set the direction and architecture for generative AI products on OpenAI and Amazon Bedrock, which my teams took from proof of concept to production in under six months across student, instructor and content workflows. They moved the numbers the business is measured on: learning outcomes up about 20 percent, drop-out down 20 percent and NPS up 25 points.
  • Built one foundational learning platform across four converging product lines using the strangler pattern, with unified data, MLOps, a shared generative AI toolkit and reusable API-driven building blocks adopted across the business.Four Platforms, One Lakehouse7 min
  • Owned the AI product roadmap, prioritized use cases by business value, and carried revenue ownership for the MathXL product line.
  • Drove modernization, cloud transformation and automation delivering $23M in annual operating cost reduction while holding 99.95 percent availability on Tier 0 platforms, including a 26TB SQL estate moved to AWS in under six months with zero downtime.I wrote this up for AWSaws.amazon.com
  • Ran large-scale A/B experimentation and model evaluation on scalable AWS data pipelines, and directed an AWS Neptune knowledge graph powering the recommendation engine, linking learners, content and outcomes.
  • Brought a major product suite back in house from vendor operation, standing up the internal team and the operating model to run it. Extreme optimization of what the platform actually needed cut the contractor base by 95 percent with zero operational impact, removing a recurring vendor cost and returning control of the roadmap.
  • Held the multi-year roadmap by managing the balance of permanent engineers and augmented capacity, using partners and staff augmentation where speed or a temporary skill gap justified it and converting to in-house ownership once the capability was established.
  • Established an architecture and AI governance portal on Backstage with Amazon Q, and partnered with Legal and Compliance on responsible AI and on WCAG 2.2 AA accessibility, which in an education business is an obligation rather than a preference.
  • Mentored technical leads to make sound calls without me, sat in architecture, code, security and operational reviews, and arbitrated the decisions that were stuck, including several outside my own group.
  • Led the change management that made the consolidation land, because a converged platform only pays back once four organizations agree to stand on it.
Senior Principal Architect Shutterfly Santa Clara, CA Jun 2015 to May 2020

Consumer photo and personalization platform at 94M active profiles, with Q4 peak as the defining constraint.

  • Led the conversion of a monolith to microservices using the strangler pattern, carving out eight core domain-driven services with explicit bounded contexts and choosing a purpose-fit datastore for each rather than one database for all of them. Ran the migration against the AWS Well-Architected Framework, with zero downtime on Tier 0 systems in nine months, saving $20M a year across backend, web and mobile.Strangling the Monolith12 min
  • Made the business case for consolidating the Identity and Access Management platform, secured the investment, and led the architects and principal engineers directly through delivery, onto OIDC, OAuth and JWT with AWS Cognito for 94M active profiles, enabling stateless high-performance applications and cutting authentication latency in half with no downtime.Identity at Scale11 min
  • Key contributor launching Identity, mobile commerce, the eCommerce platform and the single-platform approach, integrating Salesforce Service Cloud and Salesforce Marketing Cloud and exposing capability through GraphQL.
  • Solved the thundering herd problem through traffic shaping built from standard components, and built distributed session management to sustain Q4 peak traffic and enable zero-downtime deployment, clearing tech debt that had been open for years.Surviving Peak20 min
  • Created an Architecture as a Service portal with auto-generated architecture diagrams, giving a single view with drill-down for planning and transformation.
  • Established standards and design patterns, Architecture Design Records and architecture reviews across the organization, and owned the non-functional requirements for most features.
  • Led platform evaluation for IAM, eCommerce and infrastructure, built the selection metric and ran the vendor conversations through to decision.
Principal Solutions Architect Apollo Education Group San Jose, CA Dec 2009 to Jun 2015

Built the Education as a Service platform. Joined as Senior Software Engineer, then Staff Solutions Architect, then Principal Solutions Architect.

  • Initial member of the platform team that built the Education as a Service platform from the ground up, technically and organizationally, driving requirements, problem definition and the proofs of concept that were folded back into it.
  • Major contributor making the platform L-2 multi-tenant on AWS, scaling from zero to more than 300K concurrent users within 18 months.
  • Migrated the legacy platforms onto the target platform, lowering operational cost and shortening time to market for new features.
  • Architected features with instrumentation as a core tenet, producing data in a unified format so teams could build adaptive applications, an early form of the personalization now standard in AI. Built the big data platform and analysis layer underneath it on Hadoop.The patents this became11 granted
  • Conceptualized and built a dynamic configuration management system, an aggregation framework, a scalable rule engine and a third-party integration framework, and re-architected the templating layer and admin console to be modular and extensible.
  • Implemented an OWASP-based security filter, an OAuth authentication and authorization framework on Spring and a fine-grained ACL mechanism.
  • Co-owned the non-functional requirements for the platform.
Senior Software Engineer TravelMuse Los Altos, CA Nov 2008 to Dec 2009

Consumer travel planning startup.

  • Built a unified server and client-side rendering framework on JSF, EJB and MySQL, enabling dynamic snippet rendering while meeting SEO needs and ranking higher in organic search.
  • Used that framework to redesign major parts of the site to scale better in a far shorter time frame, and built vertical search and inspiration frameworks for content discovery.
  • Integrated Facebook and Twitter, which lifted user engagement.
Member Technical Staff Kasenna Sunnyvale, CA Nov 2005 to Nov 2008

IPTV middleware and the application development kit built on it.

  • Worked on next-generation IPTV middleware and built the application development kit for set-top boxes running slimmed-down Linux and miniature browsers, on a JSP and Struts MVC backend over MySQL and Oracle.
  • Wrote the client-side JavaScript framework for the front end years before React or its contemporaries existed, and shipped it as an SDK that made applications materially more responsive on hardware with very little memory or CPU to spare.
  • Delivered end-to-end features across the full lifecycle: electronic programming guide, local and network PVR, Live TV and pausing Live TV, account management and payment gateway integration.
  • Integrated DRM with external vendors such as Widevine through a Java plugin mechanism, and led onshore and offshore teams through the full development cycle.
  • Introduced the higher-revenue features on the roadmap, taking most of them end to end.
Software Engineer Hewlett-Packard Bengaluru, India Jun 2002 to Nov 2005

Enterprise services and internal engineering tooling.

  • Built an internal ticket-management application with complex workflows and introduced visualization, plus an easy-to-use knowledge base for internal users.
  • Worked on engagement-model workflows for managing intake projects with appropriate access control, end to end, and on HP's version of TeamTrack defect management across the J2EE and Oracle layers.
Co-Founder IoTango Startup venture Founder experience

Sensor networks, edge and real-time streaming.

  • Built and deployed an AI-connected IoT platform directly with customers, from solution design through production, starting with an elderly care use case.
  • Owned the full zero-to-one loop, hands-on across the stack, from architecture and code to talking with customers and deciding what to build next.
  • Networking-heavy work across sensor networks, edge and real-time streaming.

Recognized at CEO level at three companies

Given for delivered work, not tenure.

2025 CEO recognition for the agentic AI framework, and the data and MLOps implementation Wood Mackenzie
2023, 2024 CEO and Higher Education President recognition Pearson
2019 CEO Award Shutterfly
2018 Hackathon winner, AR, VR and AI Shutterfly