Year2013 – Ongoing
RoleMachine Learning, NLP
TechnologyApache Spark/ Spark ML, Scala, Parquet, AWS EC2, AWS S3, AWS Lambda, LDAViz, MLeap
Classifying similar patents is complex
Our client, a patent litigation search platform in the risk management sector, wanted to identify patents similar to a sample. The requirement was to analyze around 15mn documents (a total of 1.5TB text data) and provide technical insights into patent portfolios.
AI / ML algorithms understand patent technicalities
We developed a Machine Learning algorithm for topic modelling using Apache Spark. Thereafter, the solution was productionized on Amazon EC2 and the data was stored in S3 Automated Spark cluster setup and teardown. The learnt model was externalized from Spark Cluster to AWS Lambda for real-time prediction. This led to a flexible and economical pipeline to process patent data to derive insights.
On-demand predictive model as a service, at low cost
The on-demand predictive model came with minimal maintenance and cost which led to better operational efficiency through automated setup and teardown.