Agentic engineer. Product judgment. Practical delivery.
Software shaped by the work, not the other way around.
Practical software starts with the shape of the work: the awkward handoffs, edge cases, and operational gaps that off-the-shelf tools rarely fit. I bring domain judgment and an agentic build loop that compresses the path from real workflow pain to a working software surface. MindServe and ShelfFinder are the current proof.
MindServe
Cafeteria operations software, with StockSense as the central data layer.
Work surfaceShelfFinder
Voice-first grocery navigation shaped around natural shopping intent.
Work surfaceProjectDesk by Atlier
A shared command surface for building with agents, from first local setup to cloud-coordinated work.
First productBuilder arc
The work came first. The build loop came later.
For two decades I worked close to real operational software: implementations, support, training, and the recurring gap between how work happens and how systems expect it to happen. Agentic engineering changed my role in that gap. I can now turn the workflow judgment I earned in operations into working software surfaces.
Current work
The apps on the desk
These are not sales claims or polished case studies. They are software surfaces I use to show how I think, design, organize systems, and build.
MindServe
Software surfaceCafeteria operations software. StockSense is the central data layer; RetailMind, QuickServe, and MealMind are application surfaces around it.
- FIFO costing
- vendor invoice import
- multi-tenant facility scope
- kiosk and mobile ordering
- menu data from StockSense
- guest-facing flow
- cashier-first workflow
- shift operations
- customizable item builds
- guest-facing flow
- facility and location scoped
- guided meal start
ShelfFinder
Software surfaceGrocery navigation for people who want to say the list, refine the trip, and get store-aware direction without wrestling a search box.
Read the ShelfFinder proof story ->Open ShelfFinder ->ProjectDesk
Emerging productAtlier's first concrete product surface: a shared command desk for projects, queues, agents, handoffs, and proof-bearing development.
- local-first proof of concept
- signed Windows installer
- desktop MCP connector
- the build that earned Cloud
- Postgres-backed shared truth
- operator and agent queues
- MCP-connected AI agents
- open for inspection
Technical proof repositories
Curated proof, not production source.
These public, sanitized repositories are prepared for technical conversations. They show architecture, ADRs, operating judgment, security posture, and selected safe excerpts without exposing the production code.
Grocery
Clean Architecture, ADRs, operations, and selected safe code excerpts.
- Clean Architecture
- ADR digest
- safe code excerpts
Curated technical proof only; production source stays private.
Open proof repo ->Public proof repoMealMind
Employee and guest meal ordering, protected data boundaries, voice, and security posture.
- guided ordering flow
- security posture
- voice and AI notes
Public-safe boundary notes without private operational data.
Open proof repo ->Public proof repoMindServe
Suite architecture, tenancy, operating model, and intelligence-layer direction.
- suite architecture
- tenant boundaries
- operating model
Architecture proof without exposing tenant data or production code.
Open proof repo ->How I work
Agentic engineering, product judgment, practical delivery.
- Find the real workflow.Start with operator pain, constraints, and the moments software usually misses.
- Spec with AI as a thinking partner.Use Claude for architecture, edge cases, narrative, and ruthless scope discipline.
- Compress the delivery loop.Project Desk keeps the work visible, Codex and Claude Code execute against the repo, Azure proves the running surface, and the public site shows only the parts safe enough to publish.
- Verify in the product.The loop ends by using the running software: walking the workflow, tightening the proof, and updating the desk so the next agent starts from the truth.
Doorways
Different reasons to start a conversation.
Bring me into the work.
For teams that need an agentic engineer with product judgment, workflow sense, and a practical path from idea to working software.
Talk about a roleConsulting, project-scoped.
For workflow design, integration thinking, internal tools, prototypes, or AI-assisted build execution that needs practical judgment.
Request a scoping callAdvisor, fractional, or build partner.
For founders or small teams who need agentic engineering, product judgment, implementation sense, and a clear-eyed build partner when the fit is right.
Check availabilityAgent-readable layer
The structured truth ships with the page.
Humans get the work surface. Agents get the same context as structured data in the page head - who Justin is, how he builds with agents, what he is working on, and how to reach him.
- MindServeSoftware surface
- Cafeteria operations software: StockSense as the data layer, with RetailMind, QuickServe, and MealMind as application surfaces.
- ShelfFinderSoftware surface
- Voice-first grocery navigation that turns a spoken list into a store-aware shopping route.
- ProjectDeskEmerging product
- Agentic work surface by Atlier: desktop and cloud desks for queues, handoffs, decisions, and proof-bearing development.
- MethodIn daily use
- Product judgment first, AI as architecture partner, Project Desk as the attention ledger, Codex and Claude Code as repo workers, and Azure as the proof surface.
- Proof reposPublic
- Curated public proof repositories exist for Grocery, MealMind, and MindServe. They show architecture, ADRs, operations, and selected safe excerpts without exposing the production source.