mejohnc.org

Product archive

The broader portfolio from my existing site. These include team projects at different stages; internal systems are described here rather than linked as public demos.

Deployed internal capability

Service Delivery

A technician workspace that routes tickets into guided automations for identity, mail, and access work.

Details

Built and deployed with the service desk team; in use across identity, mail, and access work.

  • Guided automations for common identity, mail, and access work
  • Ticket context stays attached to the run
  • Built with the people who close the tickets
Deployed internal capability

Portfolio Management

A workspace that turns identity, endpoint, and security evidence into a client book you can explain — scores, gaps, and next conversations.

Details

Shipped the collection pipeline, explainable scoring, and the client book used in evidence reviews.

  • A scoped book of clients, not a generic dashboard
  • Scores that point back to a control, a source, and how fresh the evidence is
  • Gaps lower confidence instead of looking clean
  • A governed AI teammate that drafts; source systems stay authoritative
Deployed internal capability

Deep Research / Technical investigations

An investigation assistant for incidents and technical questions across identity, endpoint, mail, and operations.

Details

Built an investigation assistant that drafts cited findings across the usual operations domains.

  • Investigates across identity, endpoint, mail, and related operations domains
  • Ticket and source context stay attached to the run
  • Findings arrive cited, ready for a person to verify
Pilot

Multimodal Enterprise Agent

The assistant people talk to in Teams. It researches, remembers, and drafts — and only acts after someone approves.

Details

Led the design and delivery of the Teams assistant, now in pilot with company-grounded answers.

  • Answers grounded in handbook, people, tickets, and approved channels
  • Research and drafts with citations
  • Durable skills, reminders, and memory
  • Specialist agents share the same context and controls
Deployed internal capability

Agentic Application OS Platform

The shell where governed agents and apps run — shared company context, one approval path for anything that acts.

Details

The shell the assistants run in today, taken from idea through scoping, iteration, and review to completion.

  • Hosts assistants and specialists behind one application shell
  • Shared company context, memory, and skills
  • One OS for the agents and apps, not a pile of disconnected chat windows
Specification

Agent and Compute Federation Platform

Control plane for which identities, models, agents, and tools may run — plus cost, audit, evaluation, and rollback. Not an end-user assistant.

Details

I helped define the federation model — which identities, models, agents, and tools may run, under whose authority, and how a run is audited, evaluated, and rolled back.

  • Routes work across Azure OpenAI, Anthropic, OpenAI, and Foundry
  • Fail-closed identity and tenant isolation
  • Records runs, cost, evals, and promotions
Prototype

DevOps Teammate

Takes approved engineering work and returns a reviewed branch, a draft PR, and green CI. The model never marks its own work done.

Details

I helped design the draft-mode delivery loop: a teammate that implements and verifies against real CI.

  • Implements approved work in a fresh clone bound to a real commit
  • Opens draft PRs and repairs against actual CI
  • Surfaces decisions in Teams instead of a chat firehose
  • Humans keep merge, deploy, and anything that spends money