ARYA six-agent recruitment workflow

Six specialised agents (sourcing, screening, scoring, scheduling, outreach, reporting) coordinated with LangGraph, served by FastAPI and deployed on Azure for a recruitment team.

ARYA is a recruitment workflow built from six specialised agents: sourcing, screening, scoring, scheduling, outreach and reporting. Each agent has one job, one set of tools and one output schema, and LangGraph coordinates the hand-offs so a recruiter can follow, pause or correct the run at any step.

Candidate profiles and job descriptions are embedded into Qdrant, which gives the screening and scoring agents a shared, searchable memory. Azure OpenAI provides the models, FastAPI exposes the workflow, and the whole system is deployed on Azure for the recruitment team that uses it daily.

The design choice that mattered most was keeping agents narrow: a six-agent graph with explicit contracts was easier to test and to explain to non-technical users than one large prompt. Source code is public on GitHub.