AI Virtual Medical Recruiter

Agents + Autonomy

One of our main areas of research is using Agents and Autonomy combined. Currently frameworks such as CrewAI and Autogen do not lean them selves for a long term agents, but more yield them selves to complex network of agents for reviewing content

The issue with multi-agent frameworks is that they do not have the idea of continually learning and objective learning in mind. And are not providing the concept of continual learning, which requires the understanding of objectives and feedback loop. Here we can equate these concepts to something like reinforcement learning, where we need to understand the actions, environment, reward and long term reward

Through the use of integrating through virtual environments and real time data agents are able to self learn much better that just having a multi-agent environment alone

This idea of the autonomous actions, means that strategies are changing over time based on the current situation a agent is in. Which leads to activities at points where specific objectives are meet. As an example, when a strategy is discovered through dialog, keeping the dialog may corrupt the core findings. So through the journey of discovery it is key to compress findings and not dialog

The decision making process that we use is based similar to that of BPMN, but instead of designing the workflow up front, the AI models are making decisions dynamically in sequence or parallel.

We have found through a long term combination of Agents for continual learning, combining third part, first part data and using heuristics for assumed decisions when your are known you able to build agents that perform dramatically better, and the agent frameworks can act as multi-agent or single shot actions for researching, improving agents in small increments