Key Takeaways From Week 4 of the AI Builders Summit — Building AI
We wrapped up the final week of our first-ever AI Builders Summit! With hundreds of people tuning in virtually from all around the world, our world-class instructors showed how to build, evaluate, and make the most out of large language models. Here’s a recap of each session from this week, and if you feel like you’re missing out, then you can still see all of these sessions on-demand.
Day 1
From Idea to Implementation: How to Self-Host an AI Agent
Amanda Milberg, Principal Solutions Engineer at TitanML
Amanda Melberg demonstrated how to self-host an AI agent using Titan ML’s infrastructure. She walked through deploying a DeepSeek model on self-hosted Nvidia GPUs to perform Reddit trend analysis. The AI agent classified and summarized GenAI-related content from Reddit, using a structured pipeline with utility functions for API interactions, web scraping, and LLM-based reasoning. She highlighted the benefits of self-hosting, such as cost efficiency, privacy, and performance optimization, and emphasized using open-source models for specific applications rather than over-relying on large, resource-intensive models.
AI Q&A + Live-Speed-Build with Matt Shumer
Matt Shumer, CEO and Co-Founder of OthersideAI
Matt Schumer conducted a live demonstration of building an AI-powered search engine prototype in real-time using AI coding assistants like Cursor. He showcased how quickly modern AI tools can generate functional applications by leveraging LLMs such as Claude 3.5 and OpenAI’s O3 Mini. Schumer provided insights on optimizing AI workflows, selecting appropriate LLMs based on task complexity, and the trade-offs between small and large language models. The session emphasized the accessibility of AI development and the increasing efficiency of AI-assisted software engineering.
Build a Data Analyst AI Agent from Scratch
Daniel Herrera, Principal Developer Advocate at Teradata
Daniel Herrera guided attendees through the process of building a data analyst AI agent from the ground up. Using a step-by-step approach, he demonstrated how to integrate AI models with structured databases, enabling automated insights generation, query execution, and data visualization. The session highlighted best practices in designing AI agents for analytics, including model selection, query optimization, and ensuring interpretability. Attendees left with a clear understanding of how AI can enhance data analysis workflows and improve decision-making in business intelligence applications.
Building AI Agent Workflows: Automating Research Papers to Podcasts
Tony Kipkemboi, Senior Developer Advocate at CrewAI
Tony Kamboy explored using agentic AI frameworks to automate the transformation of research papers into podcast-ready content. The session detailed how AI agents can extract key insights, summarize complex academic texts, and generate structured scripts for audio formats. He also demonstrated workflow automation using Koo.ai, highlighting how AI-driven knowledge extraction can enhance research dissemination. The workshop provided practical insights for AI engineers and content creators looking to streamline content production with intelligent automation.
Turning Data Into Knowledge Graphs
Alison Cossette, Developer Advocate at Neo4j
Alison Cosette presented a workshop on structuring data into a knowledge graph to enhance AI-driven retrieval-augmented generation (RAG) systems. She explained how to integrate structured (SQL, CSV) and unstructured data (documents, Slack messages) into Neo4j’s graph database to create a more context-aware AI system. The session covered graph components like nodes, relationships, and embeddings, and demonstrated the process of importing, structuring, and querying data using Neo4j’s Knowledge Graph Builder. The workshop underscored the value of knowledge graphs in improving AI explainability and retrieval precision.
Day 2
Build Your Own, Better LLM-Powered Slackbot and Save Thousands
Ivan Lee, CEO / Founder of Datasaur
Ivan Lee demonstrated how organizations can build a custom LLM-powered Slackbot to automate tasks such as summarizing Slack threads, drafting emails, and generating GitHub issues — at a fraction of the cost of Slack’s AI features. He walked attendees through setting up a Slack app, configuring API integrations, and deploying the bot using open-weight models. The session provided a hands-on guide for teams looking to optimize internal workflows with generative AI while minimizing costs.
Building AI Evaluations: DeepSeek vs OpenAI in Action
Raza Habib, CEO and Cofounder of Humanloop
Raza Habib explored the importance of evaluation-driven development for AI applications. He demonstrated how to compare DeepSeek’s latest model against OpenAI’s GPT using structured evaluations, focusing on accuracy, latency, and cost. The session introduced best practices for setting up AI evaluations, running A/B tests, and ensuring model improvements translate to better real-world performance. Attendees gained practical insights into selecting the right LLMs for their use cases.
Run DeepSeek-R1
Charles Frye, Developer Advocate at Modal Labs
Charles Frye provided a deep dive into DeepSeek-R1, a state-of-the-art open-weight language model. He demonstrated how to deploy and fine-tune DeepSeek-R1 using Modal Labs’ infrastructure, highlighting the model’s efficiency, performance benchmarks, and use cases. The session covered best practices for running inference and scaling AI workloads in cloud environments, offering developers insights into optimizing model performance while maintaining cost efficiency.
Build a RAG Application in Vespa in 20 Minutes
Piotr Kobziakowski, Senior Principal Solutions Architect at Vespa.ai
Piotr Jastrzębski led a rapid workshop on using Vespa, a high-performance search engine, to build a retrieval-augmented generation (RAG) system. He covered schema design, vector indexing, and ranking functions to optimize AI-driven search and recommendation systems. The session showcased how Vespa enables scalable and efficient RAG applications, offering a compelling alternative to traditional vector databases.
Real-World Workflows for Solving Everyday Problems with AI
Holt Skinner, Developer Advocate at Google Cloud
Holt Skinner provided a deep dive into AI agents and their evolution, from basic LLM prompting to advanced workflows using LangGraph and agent-based reasoning. He demonstrated practical AI-powered workflows for engineers, including essay generation, research retrieval, and iterative refinement. The session featured hands-on coding with Google’s Gemini models and LangChain frameworks, illustrating best practices for building AI agents that integrate with real-world applications.
Advancing GraphRAG: Multimodal Integration with Associative Intelligence
Amy Hodler, Founder | Consultant | Graph Evangelist at GraphGeeks.org
David Hughes, Principal Data & AI Solution Architect at Enterprise Knowledge
Amy and David explored the future of retrieval-augmented generation (RAG) by incorporating multimodal data, such as text and images, into graph-based knowledge retrieval. They demonstrated how GraphRAG can enhance AI’s ability to understand context, improve search accuracy, and bridge gaps in knowledge representation. The session emphasized the importance of associative intelligence in AI systems, enabling more nuanced reasoning and better decision-making.
Cloning NotebookLM with Open Weights Models
Niels Bantilan, Chief ML Engineer at Union.AI
Niels Petillen walked attendees through the process of creating an open-weight alternative to Google’s Notebook LM. He demonstrated how to fine-tune and deploy AI-driven note-taking applications, using open-source models to replicate Notebook LM’s core functionalities. The session provided practical insights into building AI-powered research assistants that extract, summarize, and organize knowledge from various sources.
What’s next?
If you missed out on any of the AI Builders Summit, you’re in luck! You can watch the entire four-week event on-demand on our Ai+ Training platform for $299, or get a subscription to the platform to see those highlights and hundreds of other videos on-demand.
You can also register your interest for our next AI Builders Summit which is currently scheduled for July!
Lastly, ODSC East 2025 coming up this May 13th-15th in Boston, MA, in addition to virtually, is the best AI conference for AI builders and data scientists there is. Come learn from experts representing the biggest names in AI like Google, Microsoft, Amazon, and others, network with hundreds of other like-minded individuals, and get hands-on with everything you need to excel in the field.
