Competitor intelligence¶

Project done at Deloitte (Client: Dr Reddy Labs)

Technologies used: Python, GCP, Vertex AI, PALM 2 API Timeline: Sep 2023 - Jan 2024

POC: Chatbot which answers questions based on an article

Problem Statement
Identify potential supply issues from competitor plants of a pharma company that can arise due to FDA inspections.

FDA inspects plants across the world that supply to the US. If any issues are found during these inspections, an FDA 483 report is generated. Based on this report and the responses, it can take up to six months for the FDA to give a warning letter which can cause supply issues from the inspected plant to the USA.

Many news articles will be published on these inspections as they take place, and the results have been based on FDA 483 reports. This can help sales managers identify future opportunities due to competitors' supply issues.

The solution steps are as follows:
1. Using Google News API to find FDA inspections on pharma manufacturing plants based on time and geography constraints
2. Different web scraping methods were used to extract text from news websites
3. Generative AI was used to extract specific information like inspection starting date, ending date, location, company name, the status of the inspection, summary of results etc
4. Extracted plant and company names were mapped with FDA-published company and location data to find the drugs that could be affected by this inspection
5. Daily run was scheduled using Vertex AI and mapped data is stored in GCP

Vertex AI's PALM API finds specific information from an extracted article. Prompts with examples were created and tested for each query. Prompts are also used to get the results in a specific format.

This information is collected and synthesised across multiple articles. This is mapped with existing data shared by the FDA which has company name, location and drugs manufactured for each location.