Engineering behind the work

Useful software.
Work you can examine.

Two sides of MGV's engineering practice: building the application people use and organizing the development work that supports it.

MGV's internal Governed AI Dispatch Gateway

Human-led engineering.
Agents with specific assignments.

Using several AI agents raises practical questions: what can each one change, how do we check the result, and what happens when a job fails? Adolph built the Gateway to manage that work.

Architecture of MGV's internal Gateway, connecting human authority and orchestration to a scoped request, permitted execution, and evidence-backed review

Open the responsive architecture explanation

Problem and decision.

AI-assisted delivery needs limits and a release decision that remain clear. The Gateway separates read-only review, limited code changes, and release authority so a completed technical step is not mistaken for approval.

What was built.

LangGraph coordinates supported workflow transitions. The Gateway sets limits on what an agent may change and records the result; Git identifies the code version.

How it was checked.

One governed workflow edited a permitted file, created a specifically authorized local commit, completed acceptance testing, and received independent review.

Current result and limit

From a limited edit to a reviewed local change.

A person decides whether the work is ready to release. Current evidence covers workflows tested in MGV's macOS environment and bounded tests.

This does not establish a client deployment, certification, or ready-to-install enterprise product.

Application engineering

DNA Health Insights.

An architecture that connects genetic information with understanding.

  • Problem. Help people move from unfamiliar genetic results to plain-language educational context.
  • Design decision. Process raw DNA in the browser and save structured findings and reports to the signed-in account.
  • Built and checked. An educational application with DNA processing, saved findings, and reports; the current application source confirms browser-local processing and account-scoped report access.
  • Current result and limit. People can explore, read further, and revisit saved information. The experience does not present diagnoses or treatment decisions.

An MGV work sample created for a family member's business, not an independent client endorsement.

See the application work sample
DNA Health Insights Command Center with findings organized along a helix and a color-coded category legend
From data to understanding. Genetic findings are paired with explanations and educational context, helping people learn more and ask informed questions. Enlarge screenshot