I found an external-search path outside the intended data boundary and restricted access pending a safer design.
ROBERT GANEY / AI SOLUTIONS LEAD
I design and build AI solutions.
I help teams work through what they need from AI, how to use it responsibly, and what it takes to build and support it. My experience includes enterprise AI architecture, governance, software development, and hands-on robotics.
I built a local application that links recommendations to their sources and requires a person to approve them. The code and tests are public.
Plus 40/40 simulation checks for an autonomous six-legged rover.
SELECTED WORK
My work in enterprise AI and robotics.
These projects show what I contributed, the decisions I worked through, and the results I can share.
AI architecture for a regulated environment
RESULT
I built the pilot and held off on expanding access after testing exposed a risk.
Testing showed that the pilot could reach an external-search path outside its intended data boundary. I restricted participant access while a safer design was worked out.
REQUIREMENTS
Give the team access to AI while meeting requirements for controlled data, user access, data residency, and support.
MY ROLE
Led architecture and delivery planning, provisioned Azure AI Foundry, configured identity and access controls, coordinated vendor approvals, and tested the participant experience.
RESULT
A working restricted pilot, vendor approvals for specialized controls and model capacity, and a signed cross-functional plan. This was a pilot, not a scaled production rollout; the supporting company records are private.
ChangeFront: source tracking and review for AI work
I built ChangeFront to track the sources behind AI-assisted recommendations and document changes. If a source changes, the work based on it needs to be reviewed again.
HOW IT WORKS
A person reviews the recommendation and its sources.
An AI client can draft through the MCP interface, but it cannot approve its own work. Each recommendation cites the source versions it used, so a reviewer can check what supports it.
TESTING
Outdated sources stop being treated as current.
The tests check source changes, outdated evidence, unauthorized approvals, and altered audit records through the service, HTTP API, and MCP interface. This is a local demo with synthetic data; the code and tests are available to run.
AI rollout, governance, and employee support
RESULT
I helped teams use AI tools and put the supporting controls in place.
Workspace administration
I implemented ChatGPT and Codex safeguards across an administrative scope of 100+ users and published onboarding and usage guidance that I used in employee support.
Rollout and migration
I deployed an internal GPT across the workspace and helped a business team move from an early ChatGPT pilot into the corporate workspace. The stakeholder confirmed the support was successful.
Usage rules and exceptions
I helped define usage rules and a leadership-approved exception process for business-critical work that standard limits would block.
Governance and vendor review
I defined the inventory, controls, owners, and success criteria for a scoped governance effort, and led vendor discussions on data residency, key management, monitoring, API governance, and contract requirements.
The 100+ figure is the scope I administered, not a count of active users or a measure of adoption. Supporting company records remain private.
Read more about the rollout and support workAutonomous robotics and AI-assisted development
RESULT
Co-developed and field-tested a six-legged autonomous rover.
I worked on the Python navigation and gait code, Arduino/C++ sensor integration, telemetry, simulation, electronics, and regression tests. I also designed and fabricated the mechanical parts. The rover uses deterministic control rather than learned perception or an AI model controlling its movement.
31/32formal traversals
40/40simulation checks
7terrain categories
3COSGC recognitions
MY CONTRIBUTION
From the CAD design to the navigation code.
I co-developed the gait and navigation software, connected the Arduino sensor hub to the Raspberry Pi, and built tools for simulation and telemetry. I owned the CAD design and custom fabrication.
AI-ASSISTED ENGINEERING / CLAUDE CODE + MCP
I connected Claude Code to the rover through MCP.
I could describe what I saw the rover doing, and Claude Code could use the relevant code and telemetry to help find possible causes. I checked those explanations through simulation or field testing. Claude helped with diagnosis; it did not drive the rover.
HOW I WORK
How I work through a project.
01Understand the problem
I work with the team to identify what they need, what is getting in the way, and what success would look like.
02Work out the design
I turn those needs into an architecture, with clear access rules, responsibilities, dependencies, and tests.
03Build what is needed
Depending on the project, that means code, integrations, guidance, or a delivery plan the team can use.
04Test it and follow through
I check the result against the original requirements and make sure the limits and support needs are understood.
WHAT I AM LOOKING FOR
Leading AI projects with engineering teams.
I enjoy working through both the technical problem and the practical questions around it: who needs the system, what data it can use, how we will test it, and who will support it. My enterprise AI and robotics work has given me experience on both sides.
I'm looking for an AI Solutions Lead role where I can help set the direction and stay involved in implementation. I'm especially interested in engineering teams working across software, systems integration, and physical AI.