Are AI agents only hype or can they really improve LLM performance and create and execute action plans? The handpicked resources below motivated me to dive deeper into AI agents which have received a lot of popularity this year.
Meet our curator of the month
Monica Rafaila is an AI strategist with a history of researching novel system verification methods and building Machine Learning prototypes as a Data Scientist. She finds herself at the intersection of Data Science, AI, Product Strategy, and Innovation. She’s also curious about how music works on the human brain and how to work and travel with her 4-year old daughter.
The article, written by a globally recognized leader in AI, was one of the first drivers for the Agentic workflow hype earlier this year. It explains how agents can enhance LLM performance by using iterative strategies such as reflection, tool use, planning, and collaboration between multiple agents. These methods help LLMs improve results by reviewing outputs, gathering more data, executing complex tasks, and working collectively. The post is followed by a set of articles detailing each strategy. Andrew draws from his personal experience and links to scientific results as further basis for his conclusions.
The presentation foresees the rise of agents as the enablers for next-generation user experience and the disruption of digital products in many industries. This is justified by the ability of LLMs to analyze goals, create action plans, and execute them, as well as the ability to run LLMs on consumer devices ensuring privacy and reliable operation. Finally, key elements of designing agentic workflow are highlighted, this time from the design perspective.
This article is published on the Prompt Engineering Guide which contains a collection of proven prompt engineering methods to enhance LLM output performance. This article motivates LLM agents, explains the main LLM agents concepts, provides an overview of LLM Agent Applications including examples, Agent Tools and Agent evaluation strategies. Although the agent concepts can be found in many other resources, this one has some advantages. It is tool agnostic and based on scientific references.
The article surveys advancements in LLM-based autonomous agents, highlighting their ability to simulate human decision-making. It is rather long but can be used as a summary and reference to different techniques of implementing the profiling, memory, planning and action modules of the proposed architecture, which can be abstracted to apply in various forms of LLM applications.
Tell us about one project you’ve worked on that you found particularly interesting or challenging.
Currently piloting a partnership with a product company to build a GenAI chatbot with shared ownership. It is interesting to change my consulting mindset to fit a product company's needs.
How do you stay updated on the latest industry trends and technologies?
Any tips for maintaining a healthy work-life balance?
Choose a company like Futurice, which values individuals and their physical and mental health.
Join us!
Senior Data Engineer
Are you ready to get your hands dirty designing, building, automating and operating data platforms and data pipelines? Find out what it's like to be a Senior Data Engineer at Futurice.
Will you be a catalyst driving business development and help our clients with your advisory capabilities? Play a lead role in steering our business portfolio.
.png?upscale=true&width=320&upscale=true&name=Monica%20(1).png)
