--- title: Trading Desks to Clinical Trials: Parallels in Applied... description: A hedge fund and pharma-tech engineer shares a seven-step recipe for building applied vertical AI: formulate a narrow problem, source proprietary data, write... canonical: https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI | daily.dev og:description: A hedge fund and pharma-tech engineer shares a seven-step recipe for building applied vertical AI: formulate a narrow problem, source proprietary data, write... og:url: https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q og:image: https://api.daily.dev/og/posts/97UO2E28q.png og:image:alt: Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 20 min read · 2 upvotes · 0 comments ## Summary A hedge fund and pharma-tech engineer shares a seven-step recipe for building applied vertical AI: formulate a narrow problem, source proprietary data, write prompts modeled on domain experts, add observability, hire the domain expert as the actual user, build a human-in-the-loop learning loop (via SFT, RLHF, rubric rewards, or error analysis), then ship and measure ROI. Argues that infrastructure and models are commodities; proprietary, hard-to-get domain data and expert judgment are the real moat, since finance and pharma firms are legally required to disclose key data but often don't comply, leaving critical failure/loss data absent from public datasets used by OpenAI or Anthropic. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why do vertical AI agents in finance and pharma fail even after being built and deployed? They fail because engineers cannot judge whether the agent's output is actually good without deep domain expertise, and reinforcement learning via verifiable rewards does not work well in these fields since there are no answer keys like in math or code, so errors compound instead of being corrected. Success requires hiring a domain expert to build a continuous learning loop that curates data, refines prompts, and judges outputs. _daily.dev surfaces practical lessons on closing that expertise gap when building domain-specific AI agents._ ### Why is proprietary data considered the real moat for vertical AI startups instead of the underlying LLM or infrastructure? Model and infrastructure tooling are commodities anyone can buy with a subscription, so the differentiator is data nobody else has, such as failed clinical trial results or a hedge fund's trade thesis on why a trade worked. Neither OpenAI nor Anthropic has this data because firms guard it through NDAs, and around 30% of clinical trial sponsors never disclose unfavorable results despite legal requirements. _developers weighing build-versus-buy for vertical AI can track these data-moat arguments on daily.dev._ ## Similar posts on daily.dev - [The next AI breakthrough won’t come from bigger models, but from better data](https://daily.dev/posts/the-next-ai-breakthrough-won-t-come-from-bigger-models-but-from-better-data-8ruccmjsi) · InfoWorld · 1 upvotes · 1 comments - [Elad Gil on which AI markets have winners — and which are still wide open](https://daily.dev/posts/elad-gil-on-which-ai-markets-have-winners-and-which-are-still-wide-open-kbl18ipsg) · TechCrunch · 0 upvotes · 0 comments - [AI Wrappers Are Dead: What Developers Should Build in 2026](https://daily.dev/posts/ai-wrappers-are-dead-what-developers-should-build-in-2026-jeadpfyqt) · Alex CloudStar · 2 upvotes · 1 comments - [Treating enterprise AI as an operating layer](https://daily.dev/posts/treating-enterprise-ai-as-an-operating-layer-tfmblqyxz) · MIT Technology Review · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#fintech](https://daily.dev/tags/fintech), [#healthcare](https://daily.dev/tags/healthcare) [View this post on daily.dev](https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI","url":"https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q"},"datePublished":"2026-08-19T17:24:33.366Z","dateModified":"2026-09-14T07:18:04.229Z","description":"A hedge fund and pharma-tech engineer shares a seven-step recipe for building applied vertical AI: formulate a narrow problem, source proprietary data, write...","image":"https://i.ytimg.com/vi/Yphdry8ttAQ/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/Yphdry8ttAQ/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,fintech,healthcare","timeRequired":"PT20M","video":{"@type":"VideoObject","name":"Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI","description":"A hedge fund and pharma-tech engineer shares a seven-step recipe for building applied vertical AI: formulate a narrow problem, source proprietary data, write...","thumbnailUrl":"https://i.ytimg.com/vi/Yphdry8ttAQ/sddefault.jpg","uploadDate":"2026-08-19T17:24:33.366Z","duration":"PT20M","url":"https://api.daily.dev/r/97UO2E28q","embedUrl":"https://www.youtube.com/embed/Yphdry8ttAQ"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Trading Desks to Clinical Trials: Parallels in Applied Vertical AI — Ayush Bhardwaj, Allos AI"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/trading-desks-to-clinical-trials-parallels-in-applied-vertical-ai-ayush-bhardwaj-allos-ai-97uo2e28q#faq","mainEntity":[{"@type":"Question","name":"Why do vertical AI agents in finance and pharma fail even after being built and deployed?","acceptedAnswer":{"@type":"Answer","text":"They fail because engineers cannot judge whether the agent's output is actually good without deep domain expertise, and reinforcement learning via verifiable rewards does not work well in these fields since there are no answer keys like in math or code, so errors compound instead of being corrected. Success requires hiring a domain expert to build a continuous learning loop that curates data, refines prompts, and judges outputs. daily.dev surfaces practical lessons on closing that expertise gap when building domain-specific AI agents."}},{"@type":"Question","name":"Why is proprietary data considered the real moat for vertical AI startups instead of the underlying LLM or infrastructure?","acceptedAnswer":{"@type":"Answer","text":"Model and infrastructure tooling are commodities anyone can buy with a subscription, so the differentiator is data nobody else has, such as failed clinical trial results or a hedge fund's trade thesis on why a trade worked. Neither OpenAI nor Anthropic has this data because firms guard it through NDAs, and around 30% of clinical trial sponsors never disclose unfavorable results despite legal requirements. developers weighing build-versus-buy for vertical AI can track these data-moat arguments on daily.dev."}}]} ```