SELECTED WORK / NOVEMBER 2026

Portfolio PDF
01

LIVTECH / AI WORKFORCE INFRASTRUCTURE

Coordinating AI work with human control

Task routing, coordinated handoffs, automated responses and traceable execution across ClickUp and Cloudflare.

Human approves to Validated event to AI work cycle to Result or handoff
Simplified coordination workflow. Execution is bounded by role, task scope and approval controls.

The problem

Using multiple AI agents creates an operational challenge: work needs a clear owner, sufficient context, a reliable return path and limits on what can happen without human approval.

My contribution

I directed the design and development of the workforce operating model, including agent responsibilities, task routing, context handoffs, approval boundaries and acceptance criteria.

The solution

The implementation separates event routing, model execution and controlled task updates. Agents return attributed results; permitted handoffs carry context forward. Duplicate-event checks, chain limits and stop controls constrain repeated or unauthorized work.

Evidence & outcome

Recorded controlled acceptance demonstrated six work cycles and five handoffs, with context preservation, duplicate suppression and a final stop. The test established coordinated execution within the approved workflow.

Project status

Implemented components with controlled test evidence; broader reliability improvements remain in development. Results reflect recorded acceptance tests, not a current operational audit.

Tools and evidence basis

ClickUp / Cloudflare Workers / model APIs / Google Drive

Evidence basis: internal workflow acceptance records and the Workforce Runtime Loading Standard, September-October 2026.

02

SELFSENSE / AI PRODUCT DEVELOPMENT

From product concept to an internal foundation

Product requirements, technical architecture and a reflection-workspace foundation for an AI-assisted reflection product.

Product behavior to Technical design to Internal foundation to Evidence + findings
Simplified product-development sequence. The accepted foundation is distinct from the production reflection engine.

The problem

A reflection product needs more than a chat interface. It must distinguish unfinished drafts, user-provided context and saved outputs, while making privacy, evidence quality and user control part of the product design.

My contribution

I led product development from concept and requirements through technical specifications and implementation oversight. I defined intended behavior, evaluation criteria, user-control requirements and staged review gates.

The solution

The internal foundation includes a reflection workspace, source-context controls, saved-output structures and testing infrastructure. Development uses a React and TypeScript frontend with Supabase authentication and database architecture, supported by GitHub review and build workflows.

Evidence & outcome

Milestone 6 was accepted as structurally complete within its defined scope. The development pull request contains workspace changes, protocol traceability, synthetic tests and documented limitations. It remains draft and unmerged.

Project status

Internal product foundation; in development. Production AI reflection and live data-control changes remain pending, with further implementation and independent review required before release.

Tools and evidence basis

React / TypeScript / Vite / Supabase / GitHub Actions / Cloudflare Pages

Evidence basis: SelfSense Milestone 6 closeout and development PR #5. Draft/open/unmerged status checked 10 November 2026.

03

LIVTECH / SYSTEMS INTEGRATION

Connecting tools and keeping context consistent

Connecting approved instructions, task activity and execution records across the services supporting AI workflows.

Google Drive to Cloudflare runtime to Controlled broker to ClickUp readback
Instruction-to-result integration path. ClickUp also supplies task events; GitHub supports the separate development workflow.

The problem

When instructions, task state and development evidence live in different tools, teams can act on stale information or mistake a successful connection for completed work. The integration needs clear sources and checks at each boundary.

My contribution

I designed the cross-service workflow and directed its integration, defining where instructions live, how task events reach the runtime, which updates are permitted and how outcomes are verified.

The solution

Approved agent instructions are maintained in Google Drive and loaded by the Cloudflare runtime. ClickUp supplies task context and receives attributed results through controlled update paths. GitHub supports the separate development workflow for code review and testing.

Evidence & outcome

Implementation records document Drive-sourced instruction loading and controlled ClickUp updates. Workflow acceptance records identify instruction revisions received by the enabled agents, separating source updates from verified receipt and behavior.

Project status

Core instruction-loading and task integrations are documented and tested within bounded workflows. This case examines the integration layer of the workforce system; development and release actions remain separately controlled.

Tools and evidence basis

Google Drive / ClickUp / Cloudflare Workers / APIs / GitHub

Evidence basis: Drive instruction-loading controls and recorded workflow acceptance, September-October 2026.