Building Agents Right - MLOps Community
The MLOps Community is back in January for our first community-organized meetup of the year. This time, we'll be talking about nuances in getting agents to work right.
NOTE: REGISTRATION CLOSES MONDAY JAN 20th FOR BUILDING SECURITY.
First up we have Sam Partee, founder of Arcade AI - who will talk about building tools that personalize user experience aka interact with tools in production. Often we see tutorials of locally implemented tools. In practice, there are authentication and authorization considerations to consider when you're dealing with OAuth2 privileges and API keys. Sam will shed light on the challenges and how connecting with authed tools can help.
We also have Shreya Rajpal, founder of GuardrailsAI - Guardrails are a commonly adopted piece of infrastructure within modern GenAI applications — you can use them to prevent your AI applications from responding to off-topic requests, generating hallucinations and leaking sensitive information. How well do guardrails typically work at protecting LLM applications though? In this talk, Shreya will dive into an extensive benchmarking effort to measure the performance and latency of common open-source and closed-source guardrails and share interesting findings.
Next, Simon Farshid—founder of assistant-ui (YC W25)—will share how to design user interfaces that empower users to intervene, steer, or approve AI agent actions. By incorporating human feedback into AI-driven workflows, you can significantly boost agent reliability and transparency. He will explore practical strategies for implementing human-in-the-loop patterns and demonstrate live coding examples using LangGraph and assistant-ui.
Community Speakers: In 2025 we're also inviting community regulars to speak and share about what they've been working on.
We'll have Sri Raghu Malireddi speak about some of his work on inference optimizations. His expertise lies in bridging the gap between cutting-edge AI research and scalable deployments, enabling enterprises to adopt deep learning and LLM solutions seamlessly.
Joël Etse, Co-founder and CEO of Humiris AI, will introduce the “Mixture of AI” architecture—an innovative approach that combines the strengths of generative and predictive AI to build intelligent agents that are highly accurate, secure, and efficient.
We'll also have Jonathan Bennion who has 9 years of experience in ML and DS and has been consulting on AI projects with national firms while having also taught the class on developing end-to-end agents using Core LangChain as hosted on DataCamp - will give a live demo of agents that will fail 100% and agents that are 100% destined to be awesome
We're still building the program for this meetup, but expect to learn from some of the hottest tools available to you to build your Agents in 2025.