Deep Research for Enterprise: Unlocking Actionable Intelligence from Complex Enterprise Data with Agentic AI

Abstract

Deep Research as a consumer product redefined the AI space delivering true impact to many by searching through hundreds of websites, deeply thinking through the content, and generating a comprehensive report. In the Enterprise space, we are inundated with vast amounts of data from a multitude of diverse sources. Deeply processing and analyzing this information to uncover meaningful patterns and generate actionable insights is incredibly impactful, challenging and can be transformative to industries. 

This talk will walk through how we created Deep Research for Enterprise at Glean, detailing our approach to challenges such as designing agentic systems, complex and diverse data sources, evaluation complexity, and scalability.

Main Takeaways

  1. How a complex Agentic AI system like Deep Research is built from start to finish.
  2. Strategies for handling the challenges of complex and diverse enterprise data.
  3. Building a robust framework for automated evaluation of agentic systems.

Interview:

What is the focus of your work these days?

The majority of my time is spent on the end-to-end cycle of creating enterprise-grade Agentic AI systems. My work focuses on designing and building novel agentic architectures and then rigorously analyzing their performance through both qualitative and quantitative data to drive continuous improvement in quality and reliability.

What was the motivation behind your talk?

While the potential of Agentic AI is incredible, my experience in the trenches building Glean's Deep Research has made one thing clear: creating these systems for the enterprise is incredibly challenging. We've had to navigate a host of complex problems, from core architecture and reliability to the nuances of building user trust. My motivation comes from being at the forefront of a new paradigm in how knowledge work gets done, and I wanted to give this talk to share the details of this journey. My goal is to provide a realistic, hands-on look at what it truly takes to build a powerful AI agent, sharing the specific, hard-won lessons we learned along the way.

Who is your session for?

ML Engineers, Engineering Leads, Data Scientists, Technical Leads, etc.


Speaker

Vinaya Polamreddi

Staff ML Engineer; Agentic AI @Glean; Previously @Apple, @Meta, and @Stanford

Vinaya Polamreddi is a seasoned ML Engineer with 8 years of experience in building impactful AI products and doing AI research. She is currently at Glean where she is the tech lead for Deep Research. Previously, she built multi-modal foundational ML capabilities at Apple to launch the Vision Pro (Apple’s 1st MR headset). Prior to that, she worked on computer vision modeling at Meta’s AI Research org, and she got her Masters in AI from Stanford.

Read more

From the same track

Session

Dynamic Moments: Weaving LLMs into Deep Personalization at DoorDash

Tuesday Nov 18 / 10:35AM PST

In this talk, we’ll walk through how DoorDash is redefining personalization by tightly integrating cutting-edge large language models (LLMs) with deep learning architectures such as Two-Tower Embeddings (TTE) and Multi-Task Multi-Label (MTML) models.

Speaker image - Sudeep Das

Sudeep Das

Head of Machine Learning and Artificial Intelligence, New Business Verticals @DoorDash, Previously Machine Learning Lead @Netflix, 15+ Years in Machine Learning

Speaker image - Pradeep Muthukrishnan

Pradeep Muthukrishnan

Head of Growth for New Business Verticals @DoorDash, Previously Founder & CEO @TrustedFor, 15+ Years in Machine Learning

Session

From Reinforcement Learning Enhanced Image to AI Collection Generation: How Pinterest Cracked the Code on Content Discovery

Tuesday Nov 18 / 03:55PM PST

his talk presents Pinterest's journey in deploying AI at massive scale, from using Reinforcement Learning to create images to building 

Speaker image - Faye Zhang

Faye Zhang

Staff Software Engineer @Pinterest, Tech Lead on GenAI Search Traffic Projects, Speaker, Expert in AI/ML with a Strong Background in Large Distributed System

Session

Automating the Web With MCP: Infra That Doesn’t Break

Tuesday Nov 18 / 02:45PM PST

AI agents are only as strong as the infrastructure beneath them. In this talk, we’ll walk through the architecture behind Browserbase’s model context protocol (MCP), built to support stateful browser automation at scale.

Speaker image - Paul Klein

Paul Klein

Founder @Browserbase, previously Director of Self-Service & Engineering Manager @Mux, Co-Founder & CTO @Stream Club, Technical Lead @Twilio Inc.

Session

Engineering at AI Speed: Lessons from the First Agentically Accelerated Software Project

Tuesday Nov 18 / 01:35PM PST

Claude Code is the first developer tool built specifically to maximize AI development velocity.

Speaker image - Adam Wolff

Adam Wolff

Engineer and Individual Contributor to Claude Code @Anthropic, Previously @Robinhood, @Facebook

Session

Unconference: AI Engineering that Delivers

Tuesday Nov 18 / 05:05PM PST