We are focused on bridging the expertise-barrier, and making it feasible for generalist engineers to harness High Performance Computing like they've never been able to before.

Our GURU engineering AI assistant will enable individuals to perform an array of expert tasks — served by specialist capability 'agents' which will increase their competitiveness, and make a radical scaling up of the implementation of energy efficiency measures more feasible.

guru watch tech video

Machine Learning alone will not address the engineering expertise barrier

DARPA recently identified the need for "approaches in Artificial Intelligence (AI) that incorporate prior knowledge, such as known physical laws, to augment sparse data and to ensure robust operation"

And, they've called for "AI architectures, algorithms and approaches that "bake in" the physics, mathematics and prior knowledge relevant to an application domain in order to address the technical challenges in application of AI in scientific discovery, human-AI collaboration"

The need for hybrid-intelligence

Similarly, François Chollet of Google has called for such an approach, and represents one of a number of researchers who consider hybrid intelligence to be the solution.

"AI in the future - meta-learning systems that, given a new task:
-Quickly assemble a program that solves the problem
-Using both geometric modules (pattern recognition, intuition) and symbolic modules (reasoning, abstraction)
-Taken from a global, ever-growing library of reusable modules"

"the next wave of tooling in AI" "five years out."

This is the motivation for GURU

Platform watch tech video

GURU runs a client-server connection:

...from the user's device to services, HPC, and the 'Agent Society' containing libraries of agents that perform individual tasks. An engineering workflow is accomplished by running a series of agents collaborating together.

Each individual agent represents a hybrid:

...of a 'symbolic' rules-based system (where known best-practices, physics, mathematics, are baked-in) AND a 'geometric' Machine Learning based system which is invoked when problem-specific choices are made.

The GURU platform enables custom agents to be built in practical timeframes with sparse datasets, through methods we've developed to automate rules ingestion, and selecting and building/refining Machine Learning models.