One of the “Big Three” global consulting firms wants to automate the manual process of creating Phase 0 Due Diligence (P0DD) report that contains the analysis on how a company/product performs against its competitors. The objectives were to accomplish the following from the reviews written about the company/product 1. Predict Key Purchase Criteria (KPC) of a company/product 2. Predict the customer’s sentiment about the company/product 3. Set up several ML pipelines on AWS to build, deploy, monitor, and update models 4. Automatically generate a presentation with all the analysis provided by the KPC and sentiment models.




Deep learning models are built for predicting KPCs and sentiment for several industry verticals. The solution includes the following components:

  • Key Purchase Criteria is identified for new industry domains using topic modeling pipeline
  • Reviews are sampled through active learning pipeline for tagging (human in the loop) for the given domain
  • Automatic model selection pipeline builds and selects the best model (out of two deep learning models and one ensemble machine learning model) for the provided industry domain with an accuracy > 90% for both KPC and sentiment
  • Prediction pipeline deploys the best model in production and provides predictions
  • The production model is monitored through the monitoring pipeline
  • Retraining pipeline updates the model in production if there is performance degradation


The ML pipelines built for automatic model selection and deployment have helped in expanding the generation of the due diligence report parallelly on several industry domains in less time (earlier the process was completely manual)

The manual process of generating the report is now semi-automated, which results in reducing 75% to 80% of the manual effort by the analysts

The deck is generated automatically with all the analysis about a company/product so that the analyst can review and edit it accordingly

Our solution helped automate the due diligence report generation which helped the analysts to work on more opportunities at the same time, thereby increasing their business