Selected enterprise work
Retail Robotics & Planogram Mapping
The Home Depot | Zippedi integration
I developed the matching layer that connected imagery captured by autonomous retail robots to The Home Depot's representation of a physical store. The work translated camera observations into aisle, bay, shelf, planogram, and expected-product context.
Physical storeShelf imageryCamera observations and location metadata
My primary areaLocation matching layerNormalize, reconcile, select, and validate
Store modelPlanogram contextAisle, bay, shelf, position, and expected products
The problem
Three systems described the same shelf differently.
A camera could record what it saw. A planogram could describe what products were expected and where. Store-layout data could describe the building's physical organization. For the imagery to become useful, those representations had to resolve to the same real-world location.
My work sat at that boundary: determining where an observation belonged in the store and associating it with the correct section of the expected layout.
My contribution
From camera frame to planogram position.
- Mapped robot-generated shelf imagery to The Home Depot's store-layout domain.
- Connected observations to aisle, bay, shelf, and planogram positions.
- Reconciled location information across vendor and internal representations.
- Linked physical shelf context to the products expected at that location.
Conceptual workflow
Turning visual observations into store context.
This public-safe diagram shows the problem domain, not The Home Depot's proprietary architecture or implementation.
- 01Capture
An autonomous robot records shelf imagery and location metadata.
- 02Normalize
Location inputs are translated into a common store model.
- 03Match
Each observation is associated with a candidate planogram region.
- 04Contextualize
The image gains aisle, bay, shelf, and expected-product meaning.
Engineering perspective
What this work demonstrates.
Domain translation
Connected an emerging vendor platform to an established enterprise retail model.
Physical-digital mapping
Resolved digital observations against the structure of a real, changing environment.
Data reconciliation
Created useful context where identifiers and location assumptions did not naturally align.
Operational usefulness
Focused integration work on making robot-derived information meaningful to store systems.
Public case-study boundary
Enterprise work, responsibly presented.
This work was completed as a Home Depot employee. The description and diagrams intentionally omit proprietary source code, internal architecture, store data, implementation details, performance metrics, and confidential operational information.
Public background