Cost and Performance Optimization for AI based workloads

Ankit Virmani is an accomplished software engineer known for his expertise in optimizing Linear Programming codebases, particularly in the context of setting optimal generic drug prices for a national Pharmacy Benefits Management (PBM) business. His work involves managing contracts for over 100 insurance providers and analyzing more than one million pharmacy transactions daily. Ankit's revamp of calculations and implementation of code optimizations led to a remarkable reduction in processing time from days to mere minutes, all while enhancing accuracy.

To further enhance system performance, Ankit re-architected workflows to operate on DuckDB, a columnar database that leverages vectorized processing and parallel execution. This transition facilitated rapid, math-intensive pharmacy claims analysis, allowing DuckDB to price all drug families in under five minutes compared to the hours required by the original Python code. Additionally, Ankit employed Numba for just-in-time compilation, effectively eliminating interpreter overhead and accelerating computations.

His optimizations resulted in significant annual savings of over $25 million for the pharmacy, ensuring price match adherence and minimizing contractual violations. Ankit's improvements also mitigated risks associated with inaccuracies, guaranteeing lower prices for consumers in need of essential medications. His initiatives paved the way for customizable pharmacy benefit designs and value-based contracts, driving the adoption of agile infrastructure and leading to a company-wide migration to Google Cloud Platform (GCP).

By combining mathematical optimization techniques with scalable cloud-native software and best practices, Ankit delivered a transformative pharmacy pricing modernization initiative that not only reduced costs but also provided the reliable data necessary for crafting competitive drug pricing strategies.