Artificial Intelligence – Shabaz Basheer Patel, Anand Kiran Sampat, Lotus Ai LLC

Abstract for “Methods and systems for an end to-end artificial intelligence workflow”

“In general certain embodiments of this disclosure provide methods and systems that enable reproducible processing and scaleable deployment on a distributed network. This method involves building a machine-learning model, training it to produce multiple versions of that model, tracking those versions to create a change facilitator tool, sharing the tool with one or more devices so each device can reproduce the plurality versions of this model and then generating a deployable machine learning version through repeated training.

Background for “Methods and systems for an end to-end artificial intelligence workflow”

Modern computing systems are getting smarter every day. These smart and connected applications use data, machine learning and artificial intelligence models to operate. Although it is possible to build these models today, the process can be difficult due to the lack of standard technologies. It is difficult to track work in machine learning models manually. Manual workflows can be difficult to manage: reverting back, maintaining versions, sharing versions to ensure reproducibility and seamless horizontally-scalable deployment. A better workflow is needed to track and deploy artificial intelligence models in one place.

“In general certain embodiments of this disclosure provide methods and systems that enable reproducible processing and scaleable deployment on a distributed network. This method involves building a machine-learning model, training it to produce multiple versions of that model, tracking those versions to create a change facilitator tool, sharing the tool with one or more devices so each device can reproduce the plurality versions of this model and then generating a deployable machine learning version through repeated training.

Summary for “Methods and systems for an end to-end artificial intelligence workflow”

Modern computing systems are getting smarter every day. These smart and connected applications use data, machine learning and artificial intelligence models to operate. Although it is possible to build these models today, the process can be difficult due to the lack of standard technologies. It is difficult to track work in machine learning models manually. Manual workflows can be difficult to manage: reverting back, maintaining versions, sharing versions to ensure reproducibility and seamless horizontally-scalable deployment. A better workflow is needed to track and deploy artificial intelligence models in one place.

“In general certain embodiments of this disclosure provide methods and systems that enable reproducible processing and scaleable deployment on a distributed network. This method involves building a machine-learning model, training it to produce multiple versions of that model, tracking those versions to create a change facilitator tool, sharing the tool with one or more devices so each device can reproduce the plurality versions of this model and then generating a deployable machine learning version through repeated training.

Click here to view the patent on Google Patents.