How LinkedIn's AI hiring assistant transforms recruitment

At VB Transform 2025, LinkedIn’s chief AI officer Deepak Agarwal and senior staff software engineer Tejas Dharamsi shared insights into the company’s successful deployment of an AI hiring assistant. This agent, already live and operational, is transforming how recruiters source and engage with candidates by leveraging a multi-agent system and natural language interfaces.

The Multi-Agent Approach

LinkedIn’s AI hiring assistant relies on a collaborative network of specialized agents:

• A supervisor agent orchestrates tasks and communicates with human users.
• Sourcing agents identify potential candidates from recruiter search stacks.
• Intake agents refine qualifications based on user feedback.

"This is not a demo product," Agarwal emphasized. "It’s saving recruiters time so they can focus on nurturing candidates."

Training and Deployment

The system is trained through:

• Fine-tuned domain-adapted models for efficiency.
• Reinforcement learning and continuous user feedback.
• Experiential memory to retain context from interactions.

Dharamsi highlighted the importance of latency and reliability, with rigorous testing, red teaming, and risk assessments before deployment.

Key Features

1. Natural Language Interface: Recruiters interact in plain language, not keywords.
2. Personalization: Adapts to user preferences and learns from behavior.
3. Scalability: Standardized "Lego blocks" allow engineers to build and iterate quickly.

Challenges and Solutions

• Cost Control: Ensuring AI agents deliver value without excessive expenses.
• Privacy and Trust: Maintaining data security and consistent outputs.

"We want to provide more value to the user," said Dharamsi. "Recruiters want to focus on hiring, not searches."

For more on LinkedIn’s AI innovations, visit VB Transform.