Principal Technical Architect. Enterprise commerce on Adobe Commerce and AEM Edge Delivery Services. Edge-platform work on Cloudflare. Agentic AI on Adobe Agent Orchestrator, plus independent agent infrastructure: Felix (harness), Memoturn (memory), Fold (runtime).
I'm a Principal Technical Architect at Adobe. For Fortune 500 clients I architect and ship enterprise commerce on Adobe Commerce and AEM Edge Delivery Services — including one of the first production EDS plus Commerce B2B integrations — and increasingly the agentic layer on top of it via Adobe Agent Orchestrator. I know how Commerce systems fail, how to migrate them, and how to build teams around them.
Outside client work I build agent infrastructure across three layers: the harness (Felix, a managed-agents harness on Cloudflare Workers with manifest-defined agents, swappable session strategies, and an agentic commerce layer), the memory (Memoturn, a Rust agent-memory database with typed memories, supersession, and hybrid recall), and the runtime (Fold, a Go platform that runs any container as a Firecracker microVM with scale-to-zero — built for AI agents and MCP servers). Earlier iterations of this work ran on AWS Bedrock, AgentCore, and LangGraph. The list below is the live record; the chatbot below it answers questions about it.
8+ YEARS · ENTERPRISE COMMERCE · EDGE PLATFORMS · MULTI-AGENT AI WORKFLOWS · ZERO-DOWNTIME MIGRATIONS · COMMERCE FAILURE MODES & HOW TO FIX THEM.
Delivered one of the first production AEM Edge Delivery Services plus Adobe Commerce B2B integrations, for a Fortune 500 company. Ran a webinar on Commerce with Edge Delivery that became a reference for Adobe's enterprise teams.
Led migration of a global beverage company from Adobe Commerce on-prem to Cloud, with zero downtime. Built the architectural review process for enterprise Commerce implementations.
Led design and implementation of Adobe Commerce for a leading activewear company. Spearheaded Adobe Scene7 integration to optimize product imagery, improving visual merchandising and site performance.
Managed agents harness on Cloudflare Workers, shaped after Anthropic's Managed Agents architecture. Manifest-defined agents compile into runnable agents exposed over OpenAI-compatible chat, A2A, MCP, and REST/SSE — with Felix Commerce, an agentic commerce layer (Agentic Commerce Protocol merchant endpoint, approval-gated Stripe checkout, per-brand D2C storefronts, B2B quote-to-cash), on top. Three swappable seams: session strategies (full replay, windowed, summarizing, semantic), pattern and model-provider registries (ReAct, Deep, Router, Parallel, GroupChat, Reflect, Plan-Execute across Anthropic, OpenAI, and Workers AI), and transport-labelled tool executors (local, MCP, A2A, container, sandbox, browser, queue). Includes HITL approvals, append-only audit log, eval harness with golden datasets, guardrails, and A2A federation. Evolved from an earlier AWS Bedrock + LangGraph implementation. Docs at docs.felix.run.
Developer platform that runs any Docker image as a Firecracker microVM — scale-to-zero with wake-on-request, block volumes, a WireGuard 6PN private mesh, and OpenTelemetry-native observability (including GenAI token/cost spans). Built to be the runtime for AI agents and MCP servers. Four Go binaries — CLI, control plane (REST API, scheduler, IPAM, identity/quotas), per-host agent, and edge proxy with wake-on-request — over a cloud-agnostic provisioning interface for AWS and GCP, plus a wasmtime fast tier for near-0ms function workloads. Idle apps snapshot to zero machines and restore on the first request; triggers (cron, webhook, queue) wake work at $0 idle. Hosts MCP servers and is itself operable from any MCP client.
Agent-memory database: typed memories (facts, events, instructions, tasks) with supersession — newer facts replace older ones on the same topic with history preserved — idempotent ingest, and hybrid recall (keyword + topic + vector, rank-fused). Memory profiles can be checkpointed, forked, and rewound. Rust data plane embedding libSQL; each profile is its own tiny database that provisions in milliseconds and hibernates to object storage at near-zero cost when idle. Working prototype with 67 integration tests, true multi-node distribution (ownership, forwarding, failover, fencing) against real etcd, and a Helm chart proven on Kubernetes. Exposed over HTTP, MCP tools, and SDKs.
Toggl-like time tracking app on Cloudflare Workers, Hono, React 19, and Shadcn UI.
Configures granular emoji control across AI coding assistants. Stops emojis from cluttering documentation, console logs, CLI tools, and commit messages. A tool for engineers who want their machine output to stay legible.
Org for a suite of JUCE audio plugins for ambient music producers: Cave Delay, Cave Reverb, Cave Chorus, Cave Granular, plus a shared GUI module and licensing system. Built alongside client work, exploring DSP and real-time audio processing in C++.
Reasoning engine for eCommerce intent classification and resolution. Decomposes compound intents, resolves conflicts, outputs structured action plans. Not a simple classifier: it handles ambiguous customer intent when multiple signals conflict.
Hierarchical Durable Objects on Cloudflare Workers, with infinite nesting of manager/agent relationships through URL paths. Explores scalable real-time collaboration patterns at the edge.
Multi-agent AI system for Arizona adventure planning with 17+ specialized agents. Orchestrated multi-agent workflows for complex, multi-step planning tasks.
Production-ready Hono REST API with Bun runtime and AWS Lambda deployment. Reference implementation for serverless REST APIs with modern tooling.
Production-ready RAG system on Cloudflare Workers and Hono. Document processing, semantic search, AI-powered query responses. Reference implementation for retrieval-augmented systems at the edge.
2nd Place: Adobe AI Summit & Hackathon
We built a content generation plugin for Adobe DA.live that uses Adobe Commerce catalog data, multi-agent workflow using Langchain domain agents, and LLMs to generate AEM-ready product content. Features context-aware generation, real-time WebSocket streaming, batch page updates, and Adobe Firefly integration.
It will tell you what I've worked on and what I haven't.
I read Blake's record. Ask me what he's worked on, where, and with what.
The record is maintained as a record, not a marketing document. What's listed is what happened.