Work

Ideas into working systems

Selected work in data platforms, semantic modeling and AI-assisted analytics. Each example explains the problem, my contribution and what the system made possible.

Professional workPersonal builds

Reference architecture / Analytics engineering

Governed Analytics for BI and AI

Extending the Reporting Agent approach with explicit ownership, tested data contracts, and shared metric definitions.

Reference design: A proposed extension, not a separate production implementation.

Explore the design

Professional work

Latest professional project

Reporting Agent

Problem

People needed a practical way to ask business questions and create dashboards without rebuilding metric logic for every request.

Outcome

I built the first version of an MCP integration connecting an AI assistant to a semantic layer and warehouse-querying tools. People could use it to create analytics artifacts and dashboards.

01 / Agree

Shared metrics

Define what the business question means.

02 / Model

Semantic layer

Make the metric’s grain and supported segments explicit.

03 / Connect

Assistant integration

Give the assistant a way to request metrics and use supporting tools.

04 / Review

Check the result

Check whether the answer uses the right definition and answers the question.

Shared definitions connect business questions with analytics artifacts.
Approach & decisions
  • I worked with business and analytics partners on shared metric definitions, including the level of detail each metric represented and the segments it supported.
  • The integration gave the assistant two ways to work: requesting metrics through the semantic layer and using warehouse-querying tools for other questions. Those paths need to be treated separately. A shared metric definition establishes meaning; it does not replace access controls or review of a query.
  • The Reporting Agent built on the semantic-layer work below. My contribution was connecting that foundation to an assistant workflow people could use.

Data architecture

Shared metrics for analytics and AI

Problem

Teams needed to reuse business definitions across dashboards and analysis instead of recreating the logic in each output.

Outcome

I worked with business and analytics teams to define shared metrics and implement them in a semantic layer. That work became the foundation for the Reporting Agent.

Modeled data feeds shared business definitions. Dashboards and assistant workflows can reuse those definitions through a common interface. Authorization remains a separate responsibility.

One definition, multiple consumersModeled data feeds governed business definitions in a semantic layer. Dashboards consume the semantic layer; AI agents and applications access metrics through a shared interface.Modeled dataBusiness definitionsSemantic layerDashboardsAI agentsShared interfaceAI applicationsShared interfaceReusable foundationsMetrics · Ownership · Context
Conceptual view of the semantic-layer case study.
Approach & decisions
  • Part of the work was helping people understand how semantic models differ from individual transformation models. We needed agreement on what a metric meant, the detail it represented and which comparisons were useful.
  • In one collaboration, an analytics partner brought deep expertise in business measurement. I helped translate that knowledge into reusable semantic definitions. They began modeling this way, and their team started contributing.
  • That adoption mattered: the definitions became something business and analytics partners could develop together.

Analytics modernization

ELT architecture and modeling standards

Problem

Analytics delivery needed a more consistent way to move from a business request to a usable model.

Outcome

I implemented an ELT architecture, created modeling standards and documented a framework for analytics delivery.

  • ELT
  • Modeling standards
  • Analytics engineering
  • Delivery planning
Approach & decisions
  • I connected the architecture work to an engineering roadmap and analytics priorities. The standards and documentation gave the team a common starting point for new models and delivery decisions.

Revenue systems

Subscription management and revenue systems

Problem

Self-service customer journeys needed systems that connected subscription management, product usage and account information.

Outcome

I built subscription-management functionality and architected self-service GTM systems connecting product usage with account intelligence.

  • Subscription management
  • Revenue systems
  • Business applications
  • Product-led growth
Approach & decisions
  • The work included warehouse, ELT and reverse-ELT data flows, alongside business applications. I managed the systems portfolio and worked across teams on technical priorities and delivery.

Personal project / Data, semantics & AI

CaptGabe Fitness

Built for my own training. A framework I want to grow and scale.

Actual CaptGabe Fitness dashboard showing a planned lower-body workout and nutrition targets
Actual dashboard · September 29, 2026 · Planned session and daily targets; no food logged that day. Select to enlarge.

The builder instinct

I wanted to create an app for myself, then build the framework to support its growth. The detailed expansion plan is still to come.

The feedback loop

From entries to a personal assistant.

Structured workout records feed data models. Metric definitions give ChatGPT consistent meaning when discussing progress and adjustments.

“How has my incline press progressed?”

“Where has performance stalled?”

“What should we adjust next session?”

Why the semantic layer matters

A shared definition of each metric makes comparisons consistent. The assistant can discuss training history in the context of exercises, sets, load, and planned targets.

I choose the adjustment, complete the next workout, and bring those results back into the same feedback loop.

01 / Capture

Training, nutrition & measurements

Workout entries, sets, reps, load, food logs, and body measurements.

02 / Store & model

Fitness data models

Workouts → exercises → sets, alongside food entries, measurements, programs, and planned targets.

03 / Define meaning

Semantic & metric layer

Reusable views for latest exercise performance, daily nutrition actuals, and progress against the plan.

Dashboard

See the progress

The authenticated app reads data and metric views through its dashboard API.

Assistant workflow

Ask ChatGPT

Explore fitness metrics, discuss patterns, and use training history to inform the next workout.

Close the loop ↺

Review the recommendation → choose an adjustment → train → record the next session.

Data models, metric views, and dashboard access are reflected in the app source. The dashed assistant branch represents the ChatGPT workflow; its connection is managed outside this app repository.

Personal build

This website

AI-assisted editing, version control, and automated publishing shorten the path from idea to a finished page.

Explore the build