Drug-Drug & Drug-Lifestyle Interaction Engine

The Challenge

• No standardised interaction database existed internally.
• Over 5.5 lakh raw interactions from various sources needed to be cleaned, classified and mapped to Tata 1MG's internal drug IDs across drugs and products.
• FDC (fixed-dose combination) interactions required a novel logic framework, as standard interaction tools do not account for combination drug behaviour.
• Oncology drug interactions required a separate pipeline due to their high risk, highly specific nature and the involvement of an external hospital partner.

The Approach

• Used Python and BeautifulSoup4 to scrape and structure data across 5.5 lakh interaction entries, building a phased delivery pipeline across severity tiers.
• Designed a consume type level mapping system to enable interaction checks at the granular drug-form level, not just the salt level.
• Built interaction libraries for red (life-threatening/severe), yellow (moderate), blue (mild), and green (no interaction) categories, each with severity grading, explanations, and layman friendly action text.
• Led reference mapping for ~2 lakh DDI entries and ~1,000 DLI entries to anchor every interaction to a citable clinical source.
• Collaborated with the Tata Memorial Centre to isolate and deliver 4 batches of oncology-specific interactions (~31,000 interactions) with specialised validation.
• Licensed the DDI tool to Samsung Health as part of a $175,000 commercial partnership, covering 2 lakh interactions across 5 lakh+ product combinations.

Results

Tech Stack