Inspiration
The traditional recruitment pipeline is broken. HR professionals spend countless hours manually screening unstructured resumes, leading to fatigue, unconscious bias, and slow hiring cycles. We were inspired by the concept of Agentic Business Orchestration—the idea that workflows shouldn't just follow static rules, but should utilize specialized AI agents to make cognitive decisions. We wanted to build a system where AI does the heavy lifting of parsing, evaluating, and reasoning, leaving HR to simply review the final, structured insights.
What it does
Recruitment Intelligence Agent is a fully autonomous UiPath BPMN workflow (Track 2: UiPath Maestro BPMN). It natively accepts key parameters directly via UiPath Orchestrator Start Event arguments: the jobPosition, the jobDescription , and the candidate's resume pdf (File).
It routes these inputs through a sequential pipeline of specialized agents:
Resume Parsing Agent: Takes the uploaded PDF file, extracts the unstructured text, and converts it into structured data (Skills, Education, Experience).
Recruitment Intelligence Agent: The core decision-maker. It cross-references the extracted profile against the specific JobDescription to determine a precise fitment score and generates reasoning.
Decision Recording Agent: Finalizes the AI's evaluation into a clean, structured JSON log (Accepted/Rejected) ready for HR review.
How we built it
We focused purely on building a robust agentic orchestration backend to highlight the true power of UiPath Automation Cloud.
Orchestration & Logic: We utilized UiPath Studio's BPMN designer to map the multi-agent sequence.
Execution: Hosted on UiPath Automation Cloud running via an Unattended Robot.
Triggering: We configured In/Out arguments directly on the Orchestrator Start Event to natively accept string parameters and the physical PDF file.
Note (Bonus Points): We accelerated our development process by utilizing AI coding assistants to help us brainstorm system prompts, structure the expected JSON schemas for our agents, and design the logical flow of the BPMN!
Challenges we ran into
Our biggest hurdle was Context Passing and Prompt Engineering between multiple sequential agents. When Agent 1 (Parser) extracted text from the PDF and sent data to Agent 2 (Recruitment Intelligence), we noticed that unstructured or hallucinated formats caused Agent 2 to fail its evaluation. We had to heavily refine our system prompts and enforce strict JSON-schema outputs to ensure that the "handoff" of data between the UiPath agents was seamless and predictable.
Accomplishments that we're proud of
We are incredibly proud of building a pure, headless Agentic Orchestrator. We proved that UiPath doesn't just need to be a robotic task executor; it can serve as the central cognitive brain and governance layer for complex, AI-driven business decisions—natively handling file uploads and string processing without needing a massive external tech stack to function.
What we learned
We gained deep, practical knowledge of the UiPath Orchestrator and Maestro BPMN ecosystem. We learned how to map Orchestrator Start Event arguments to handle files and strings natively, manage state and context across multiple distinct AI agents, and enforce structured data outputs (JSON) from generative models within a sequential business flow.
What's next for Recruitment Intelligence Agent
Since our core reasoning engine is fully operational, our immediate next step is building the integration layer. We plan to build a React.js dashboard and a Node.js backend that utilizes UiPath Storage Buckets and the Orchestrator StartJobs API. This will allow HR to upload bulk PDF ZIP files, have them processed via the backend, and automatically trigger this BPMN flow via Webhooks, creating a complete "Human-in-the-Loop" full-stack platform.
Built With
- agents
- orchestration
- uipath
Log in or sign up for Devpost to join the conversation.