Amigo Overview
Last updated
Was this helpful?
Last updated
Was this helpful?
Amigo is a comprehensive platform providing the cognitive architecture and orchestration framework for enterprises to build, deploy, and manage safe and reliable AI agents for demanding industries (e.g., healthcare, legal, finance). Key design considerations include performance, safety mechanisms, reliability features, and observability.
To understand the design of our system, we must discuss how complex problems are solved.
At the heart of effective agent training lies a three-layer framework that creates a coherent approach to addressing complex challenges.
The Problem Model is a comprehensive representation of the problem space to be addressed, encapsulating not just surface-level challenges but the deeper contextual understanding of the domain. Defining an opinionated stance on what the problem model should be requires domain expertise and specialized data foundations that evolve over time. Organizations define this layer by articulating what problem needs solving and establishing the boundaries and characteristics of that problem space.
The Judge exists to answer the following question: 'What does successfully solving the problem look like? Organizations must define the critical evaluation framework that embodies the strategic objectives that define when a problem is considered solved. As market conditions and problem definitions evolve, so too must the evaluation criteria. Organizations define this layer by articulating what success looks like in their specific context and how it should be measured. The AI system needs to provide the infrastructure on which clear success criteria can be defined and evaluated against at scale.
The Agent occupies the central position in this framework, serving as the dynamic problem-solver that operates within the bounds of the problem model and optimizes towards success measures as determined by the Judge. The job of the Agent is to leverage LLMs (large language models) to solve the problem in the optimal way using limited computational resources. Continuously evolving to meet the challenge at hand, the agent exists in a state of productive tension - squeezed between the requirements of the Problem Model and the expectations of the Judge.
Amigo's architecture operates on a clear division of responsibilities: your are primarily responsible for defining the problem models and judges that drive evolutionary pressure and track competitive market changes, while Amigo focuses on building an efficient, recursively improving system that evolves under that pressure. This partnership is fundamental to creating AI that is not only technologically advanced but also deeply embedded and effective within your specific operational context. As markets evolve and problem definitions shift, our partners continuously sharpen these inputs through specialized data, refined problem scopes, or updated success metrics. This collaborative approach leverages the strengths of both parties—domain expertise from our partners and technical innovation from our team—to create agents that continuously improve over time.
What makes our system unique lies in our fundamental understanding of how effective agents process information and make decisions: Memory, Knowledge, and Reasoning (M-K-R) are not separate components but are best understood as different facets of a single, interconnected cognitive problem. Their optimization is a cyclical process, and the integration bandwidth between them is critical to overall agent intelligence and performance.
The power of our approach lies in recognizing that these components form a unified system with complex interdependencies:
Memory influences how knowledge is applied and how reasoning is framed—such as when memory of a user's broken leg changes how exercise knowledge is applied and which reasoning paths are prioritized.
Knowledge and new reasoning, in turn, impact how memory is recontextualized—as when a critical piece of information about heart attack risk causes all previous symptoms stored in memory to be reevaluated in a new light.
Reasoning, while dependent on both knowledge and memory as direct inputs, also affects how they're utilized—different reasoning frameworks lead to different interpretations even with identical knowledge and memory bases.
This unified view creates a virtuous, cyclical optimization path where improvements in any one area cascade through the entire system. Optimizing memory enhances knowledge utilization and reasoning capabilities. Refining knowledge organization improves memory contextualization and reasoning paths. Strengthening reasoning processes leads to better memory utilization and knowledge application. Rather than treating these as isolated components requiring separate optimization strategies, our architecture recognizes their fundamental interconnectedness and the critical role of high-bandwidth integration between them.
