Submission type
Coded Agent with UiPath SDK
Name
Piotr Kolakowski
Industry category in which use case would best fit in (Select up to 2 industries)
Healthcare Pharma
Operations
Complexity level
Advanced
Summary (abstract)
The Calibration Dispatcher Agent plans and dispatches medical-device calibration visits across multiple clinics. It analyzes equipment status (Data Fabric), retrieves policy limits via Context Grounding, builds optimized routes with Google Maps, and runs a Human-in-the-Loop approval in Action Center. After approval, it triggers RPA via MCP to send emails, post to Slack, and create Service Orders in Data Fabric. Built with UiPath SDK + LangChain (LangGraph) + MCP, it demonstrates a full ReAct loop with explainable decisions and safe escalation.
Detailed problem statement
Scheduling calibration for audiology equipment across cities is typically manual, slow, and error-prone:
- Fragmented data & rules: device status lives in entity records, while policy limits (max hours/visits/distance, SLA for OVERDUE/URGENT) are stored in documents/spreadsheets; operators search and interpret them manually.
- Inefficient routing: technicians are assigned without consistent distance/time optimization; travel overhead inflates costs and puts SLA compliance at risk.
- Specialization constraints: not every technician can service audiometers vs. tympanometers; matching skills to clinics is done by hand and often late.
- Approval ping-pong: managers need visibility and an approval path (approved / change request / rejected); without a structured HITL loop, revisions extend lead time.
- Execution gaps: after approval, notifications and service-order updates still require manual emailing, messaging and data entry.
The result is hours of weekly planning, inconsistent adherence to policy/SLA, and limited auditability. This use case targets those pain points with an agent that can reason over data, optimize routes, keep the human in the loop, and execute the after-approval steps automatically.
Detailed solution
Architecture (end-to-end)
- Agent (ReAct) — UiPath SDK + LangChain
- analyze_equipment_status() reads Equipment/Clinics/Technicians entities from Data Fabric and classifies devices: OVERDUE / URGENT / SCHEDULED.
- get_calibration_rules() uses Context Grounding to retrieve limits: max hours, max visits, max distance.
- build_routing_plan() groups by city, matches technician specialization, calls Google Maps waypoint optimization, and produces dated routes constrained by policy.
- HITL approval — UiPath Action Center
- For each route, the agent creates an Action task that shows map + details; manager selects Approved / Changes Requested / Rejected (≤3 iterations). Decisions flow back to the agent to re-plan or proceed.
- Execution via MCP (post-approval)
- The agent triggers RPA through MCP tools:
• Email route confirmations,
• Slack summary to channel,
• AddServiceOrder entity record creation. - Integration is exposed as 3 reusable MCP tools.
- Observability & tracing
- Clear INFO/WARNING/ERROR logs and traceable steps for Reason/Act/Observe; sample key log lines included in repo.
Technology stack: UiPath SDK (Python) + LangChain, Context Grounding, Action Center, Data Fabric, MCP, Google Maps, Slack, M365.
Narrated video link (sample: https://bit.ly/4pvuNEL)
Expected impact of this automation
• Planning time ↓ 60–80% - routes and approvals generated in minutes instead of hours.
• SLA & policy compliance ↑ - constraints enforced automatically with HITL for edge cases.
• Travel overhead ↓ - waypoint optimization + city grouping reduces total distance.
• Auditability & transparency ↑ - every approval and execution step is logged and traceable.
UiPath products used (select up to 4 items)
UiPath Action Center
UiPath Apps
UiPath Automation Cloud™
UiPath Coded Agents
UiPath Data Service
UiPath Integration Service
UiPath Orchestrator
UiPath Studio Web
Integration with external technologies
Google Maps API, Slack, Microsoft 365, OpenAI
Hi there, 
Nov 5–16: Jury evaluation by 
Nov 17: Winners announced 
.
,