2025-11-09 –, Room 127
OpenAI and Gemini's Deep Research offerings are a great way to get a detailed research report on a topic.
In this beginner friendly tutorial, we’ll walk through building a simple lightweight agent workflow to perform deep research.
OpenAI and Gemini's Deep Research offerings are a great way to get a detailed research report on a topic.
In this beginner friendly tutorial, we’ll walk through building a simple lightweight agent-style workflow to perform deep research.
A Deep Research Workflow is composed of -
(a) Understanding what the user is asking for
(b) Performing web searches on the topic
(c) summarizing the findings to a report
For this workflow, we would need a LLM, agent framework, web search tool.
For this session, we will orchestrate the workflow using LangGraph, with OpenAI as the LLM and Tavilly as the web crawl provider.
Outline
- [:00 - :05] Intro
- [:05 - :10] Local Setup
- [:10 - :25] Langgraph Fundamentals
- [:25 - :40] Scoping with an LLM
- [:40 - :60] Tool Use for web search
- [:60 - :80] Summarizing the findings
- [:80 - :85] Some production implementation and evaluation
- [:85 - :90] Q/A
Links
Setup
We will use Github Codespaces.
The workshop uses OpenAi as the LLM and Tavilly as the web search tool.
During the wokrshop, we will provide a key.
If you would like to use your own setup, please follow the below links-
Takeaway
We hope that attendees will walk away armed with the knowldege on how to use a framework like LangGraph to build an agentic workflow to solve problems like deep Research.
ML Engineer at Walmart
I’m currently a full-stack machine learning engineer at Walmart E-commerce, where I get to tackle exciting challenges in the world of online retail. Before that, I was a data scientist at Bank of America, building real-time fraud detection models using deep neural networks and big data – talk about high stakes!
My research interests lie in the fascinating areas of graph embedding, neural architecture search, and fast optimization methods for neural networks. I love pushing the boundaries of what’s possible with AI.
But my passion for technology extends beyond my day job. I’m also deeply invested in two side projects:
AI-Powered Vision for IoT: I’m exploring the potential of NVidia Jetson Nano to create innovative machine learning vision applications for the Internet of Things.
ML Design Patterns: I’m developing reusable design patterns to solve common machine learning problems, making AI development more efficient and accessible.
And when I need a break from the digital world, I head to my garden. I’m an avid grower of Cayenne peppers – the hotter, the better!
My journey to AI was paved with diverse experiences. Earlier in my career, I worked on NLP-based automated evaluation of text data, gaining valuable insights into the power of language processing. I hold a master’s degree in computer science from North Carolina State University – Raleigh (graduated in Spring 2016) and a bachelor’s degree in electronics and communication engineering.