← Multimodal Sources 🕐 2 min read
Multimodal Sources

Scaling LinkedIn’s Hiring Assistant with Multi-Agent LLM Systems | VB Transform 2025

Executive summary · March 2026

> We are going to have one supervisor agent that is orchestrating all the tasks among other agents. Every other agent [...] they are good at one and only one job.

Show: Beyond the Pilot · Publisher: VentureBeat · Host: VentureBeat editorial

Episode URL: https://www.youtube.com/watch?v=XgKs0aw8ssg

Publish date: 2026-04-13
Duration: 1229.0s
Default source credibility: HIGH — Named F500 practitioners on-record with production metrics. VentureBeat editorial vetting. Treat vendor-sponsored segments as MEDIUM.

  • LinkedIn’s AI strategy uses recommendations, search, and agentic systems to connect talent, knowledge, and services, aiming to create economic opportunity.
  • The Hiring Assistant employs a multi-agent architecture where a central ‘supervisor’ agent coordinates specialized sub-agents for specific tasks like intake and sourcing.
  • Early results from the Hiring Assistant show 73% of surveyed charter customers saving over an hour of sourcing time per role, validating the agentic approach.
  • To scale development, LinkedIn created a GenAI App Platform with reusable components, enabling teams to build and deploy new agents faster and more consistently.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Deepak Argawal (Chief AI Officer) LinkedIn 08:27 12-agent-workers We are going to have one supervisor agent that is orchestrating all the tasks among other agents. Every other agent […] they are good at one and only one job. They don’t communicate with each other. All communication happens through just one agent, which is what we are calling the supervisor agent.
2 HIGH Tejas Dharamsi (Sr. Staff Software Engineer) LinkedIn 20:11 12-agent-workers We’re not just building AI tools, rather creating an ecosystem of specialized agents that work together. This requires more than just orchestrating other LLMs/LRMs.
3 HIGH Tejas Dharamsi (Sr. Staff Software Engineer) LinkedIn 20:01 07-adoption-challenges Human in the loop remains critical; agents should augment, not replace.

Per-quote detail

1. Deepak Argawal — LinkedIn (08:27)

We are going to have one supervisor agent that is orchestrating all the tasks among other agents. Every other agent […] they are good at one and only one job. They don’t communicate with each other. All communication happens through just one agent, which is what we are calling the supervisor agent.

  • Credibility: HIGH — C-level executive at a major tech company describing a specific, non-obvious architectural choice for their multi-agent system in production.
  • Topic tag: 12-agent-workers

2. Tejas Dharamsi — LinkedIn (20:11)

We’re not just building AI tools, rather creating an ecosystem of specialized agents that work together. This requires more than just orchestrating other LLMs/LRMs.

  • Credibility: HIGH — Senior practitioner articulating the strategic shift from building monolithic tools to creating a collaborative ecosystem of specialized agents.
  • Topic tag: 12-agent-workers

3. Tejas Dharamsi — LinkedIn (20:01)

Human in the loop remains critical; agents should augment, not replace.

  • Credibility: HIGH — Senior practitioner stating a core design principle for their agentic systems, emphasizing augmentation over replacement, which is a key lesson for enterprise adoption.
  • Topic tag: 07-adoption-challenges

Extracted 2026-04-13T23:40:35 via scripts/podcast_mine.py (Gemini gemini-2.5-pro).