Retail Operations Control Tower
Orders, inventory and service health in one decision dashboard.
FDE Academy · Applied AI Engineering
One client.
Eight harder problems.
Start with a retail AI API. Add enterprise data, evaluated retrieval, approved agent actions and production constraints. Then take on a new client brief.
08
connected modules
17
planned builds
01
unfamiliar-client capstone
SIMULATED CLIENT / CIRRUS RETAIL
SYSTEM VIEW
Discover
Decide
Build
Integrate
Defend
How the curriculum advances
Eight modules. Increasing stakes.
The retail platform keeps evolving. Select a module to see the engineering focus and the client decision behind it.
8 MODULES + CAPSTONE
01W01–05Ship a client AI serviceRetail intelligence API02W07–12Connect real business systemsRetail operations control tower03W14–19Ground answers in evidenceEvaluated support copilot04W20–23Let agents act safelyHuman-approved resolution agent05W24–27Orchestrate work and codeRecoverable multi-agent case desk06W28–33Survive production pressure50× traffic readiness test07W35–36Route work to the right modelInference strategy benchmark08W37–40Adapt and serve a small modelPrivate small-model service
MODULE 01 / FIRST RELEASE
One request becomes a working cloud service, with tests and a recoverable container release.
Engineering
FastAPI · LLM APIs · Docker · AWS
Consultancy
Discovery, PRD, client sign-off
CIRRUS / CLIENT BRIEF 03
RELEASE REVIEW
“Our support assistant gives the wrong refund policy.”
TECHNICAL TRACK
Improve retrieval. Test cited answers against a golden set.
CONSULTANCY TRACK
Define acceptable errors, exposure and cost per case.
Both tracks meet
Ship, revise or hold the release?
DEFEND THE DECISION
One project / two responsibilities
The code has to work. The release has to make sense.
In the same week, learners test the system and face the customer decision it creates. Both belong in a Forward Deployed Engineer's delivery.
FDE L.A.B. · Learn → Apply → Build → Review
The applied portfolio
Six builds worth showing.
Selected from 17 connected engineering projects. Each introduces a different production or client constraint.
PROJECT WORK / SELECTED
Enterprise Knowledge Copilot
Answers grounded in policy documents, with citations and retrieval tests.
Human-Approved Resolution Agent
An agent pauses for approval before executing a business action.
Agentic DSA Dispatch Engine
Graph-based task dispatch with dependency queues and concurrency limits.
Black Friday Resilience Lab
Stress-test a 50× traffic surge and rehearse service recovery.
Private Small-Model Inference Service
Fine-tune, benchmark and serve a private small language model.
- P01Retail Intelligence APIW01–W05
- P02Ship-Anywhere Release KitW02–W05
- P03Tenant-Safe Commerce Data ServiceW07–W08
- P04Legacy Orders Reconciliation PipelineW09–W12
- P05Real-Time Inventory Event EngineW10–W12
- P06Retail Operations Control TowerW11–W12
- P07Enterprise Knowledge CopilotW14–W19
- P08AI Quality & Release GateW16–W19
- P09AI Security Red-Team LabW18–W19
- P10Human-Approved Resolution AgentW20–W22
- P11Enterprise Tool & MCP GatewayW23–W27
- P12Agentic DSA Dispatch EngineW25–W27
- P13Multi-Agent Operations DeskW26–W27
- P14AI Release & Incident ConsoleW28–W33
- P15Black Friday Resilience LabW30–W33
- P16Private Small-Model Inference ServiceW35–W40
- P17Ride-Hailing Exception Desk (Capstone)W41–W44
Final engagement · W41–44
New client. Different domain. Same accountability.
The final brief moves to a simulated ride-hailing marketplace. Learners scope and deploy an exception-resolution service, then defend its architecture and handover.
MOVE / SIMULATED RIDE-HAILING CLIENT
01 / DISCOVER
Trip exceptions & disputed cases
02 / BUILD
Agent actions with approvals
03 / DEFEND
Audit trail & client handover
Capstone project: Ride-Hailing Exception Desk.
The weekly rhythm
One learning loop, applied every week.
TUE
Technical
WED
Consultancy
THU
Technical
SAT
Build & review
SUN
Selected clinics
Technology, used in context
The tools follow the engineering problem.
Application: Python, FastAPI, PostgreSQL, Docker. Agents: LangGraph, MCP, agentic harnesses. Quality: RAG evaluation, golden sets, security testing. Operations: AWS, Terraform, Kubernetes. Models: routing, LoRA, QLoRA.
Curriculum note. CIRRUS and MOVE are simulated teaching clients. Projects describe planned learner builds and do not imply commercial deployments or affiliation with named companies. Specific tools and schedules may be updated.