Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI
Ayush Bhardwaj works with AI tools in two very different industries — financial trading and drug development — and discovered that the problems are nearly identical.
The major challenge is assessing whether the AI system being built actually works well: an experienced coder can tell in a minute if generated code is weak, but no one on his team could judge the quality of a trading strategy or a drug candidate proposal. He first tried solving this by training a model to evaluate the results, but it failed because critical data is kept secret (trading funds don't want to reveal winning strategies, and approximately 30 percent of pharmaceutical companies never publish their trial results). The solution became hiring a senior expert from each field who can assess results, select the best data sources, and refine how to ask the AI system for help. The expertise and data are what create real competitive advantage — the AI model and infrastructure themselves are essentially interchangeable.
Vibekollen prepared this summary with AI from the original publication. The content belongs to AI Engineer.
