Marketing Agent

Explore an AI agent for Toast marketing, its architecture, and the challenges of live deployment, including HMAS pattern trade-offs.

HMAS AI LLM NLP ML

Overview

An AI Agent for Marketing within the Toast ecosystem. We’ll demo the agent and what it does, and we’ll discuss some of the important architecture aspects of it.

Tech stack

HMAS

HMAS (Hypermedia Multi-Agent Systems) is an open-source W3C community standard and ontology designed to align autonomous AI agents with the native architecture of the Web.

HMAS shifts the paradigm of AI coordination by moving away from closed, flat agent swarms and toward a web-native, hypermedia-driven architecture. Grounded in W3C Web of Things (WoT) standards and semantic web technologies, the hMAS core ontology defines how heterogeneous entities (including LLM agents, human operators, and IoT devices) discover, navigate, and interact with one another through web-based resources and structured affordances. By treating tools and agents as standard web resources, HMAS ensures that complex AI systems remain highly modular, scalable, and fully interoperable across distributed enterprise environments.

https://github.com/HyperAgents/hmas

View projects
AI

AI: The computational system driving human-level problem-solving (e.g., GPT-4, AlphaGo), actively transforming sectors like healthcare and finance with predictive analytics.

Artificial Intelligence (AI) is the system's ability to simulate human cognitive functions: learning, problem-solving, and decision-making. Key models like OpenAI's GPT-4 and Google DeepMind's AlphaGo demonstrate rapid capability expansion across diverse domains. This technology is actively deploying across critical sectors: healthcare uses AI for diagnostic image analysis (often achieving 90%+ accuracy), finance employs it for real-time fraud detection, and autonomous vehicles (Level 4) rely on its processing power. Global investment validates this impact: the AI market is projected to exceed $1.8 trillion by 2030 (a clear indicator of scale). Focus now shifts to responsible scaling and robust governance (e.g., data privacy, bias mitigation) to manage widespread integration.

https://aaai.org

View projects
LLM

Large Language Models (LLMs) are deep learning models, built on the Transformer architecture, that process and generate human-quality text and code at scale.

LLMs are a class of foundation models: massive, pre-trained neural networks (often with billions to trillions of parameters) that leverage the self-attention mechanism of the Transformer architecture (introduced in 2017) to predict the next token in a sequence. Trained on vast datasets (e.g., Common Crawl's 50 billion+ web pages), these models—like GPT-4, Gemini, and Claude—acquire predictive power over syntax and semantics. They function as general-purpose sequence models, enabling critical applications such as complex content generation, language translation, and automated code completion (e.g., GitHub Copilot). Their core value: generalizing across diverse tasks with minimal task-specific fine-tuning.

https://en.wikipedia.org/wiki/Large_language_model

View projects
NLP

Natural Language Processing (NLP) is the AI subfield that teaches computers to interpret, manipulate, and generate human language, powering critical applications like Siri, Google Translate, and enterprise sentiment analysis.

NLP is the core technology bridging human communication and machine intelligence: it combines computational linguistics with deep learning models to process text and speech. The process starts with tokenization (parsing language into elemental pieces), followed by syntactic and semantic analysis to determine structure, context, and intent. This capability is leveraged for high-value tasks, including automating customer service via conversational AI, performing real-time sentiment analysis on millions of data points, and enabling machine translation across dozens of languages. Enterprises using NLP have reported significant gains, such as achieving a 383% ROI over three years by streamlining operational workflows.

https://www.ibm.com/products/natural-language-understanding

View projects
ML

ML is the AI subset where algorithms automatically learn patterns from data to make predictions or decisions, replacing explicit, hard-coded instructions.

Machine Learning (ML) is an artificial intelligence subset focused on building systems that learn directly from data: it is not explicitly programmed. ML algorithms, including neural networks, ingest large training datasets to identify complex patterns and optimize a model's performance. This process allows the model to generalize and make accurate inferences on new, unseen data. Key applications drive major industry functions: recommendation engines (e-commerce), fraud detection (finance), and computer vision (autonomous vehicles) all leverage ML to improve efficiency and automate decision-making at scale.

https://www.ibm.com/topics/machine-learning

View projects

Marketing Agent

Related talks