Healthcare services · Clinical drug information
InpharmD
An Atlanta service that answers clinicians’ drug questions with written evidence summaries. A named pharmacist signs every answer; software does the searching and the first draft. It has named health system customers, and no independent evaluation of how accurate its answers are.
A sentence arrives on the monitor
In 2018 a public radio reporter sat in a small glass office at Atlanta Tech Village and watched a line of text appear on a monitor in the corner. Somewhere in metro Atlanta, a doctor had just asked a question about a drug. Neil Patel, a drug information pharmacist and one of a handful of people then on the payroll, would either answer it himself or push it to a university drug information center.
“It could be a simple dosage question, or it could be something that I would have to look up, a complete disease state that I’ve never heard of,” Patel told WABE.
On the wall behind him, in large letters: Mercer University. The private university put up the first money, which Advani said was a first for the school.
A fifth of the working day
Hospitals hire clinical pharmacists to get the right drug to the right patient at the right dose. A large slice of that job is fielding one off questions from physicians, and answering them means going to the literature. “20% of their days are just spent on PubMed, looking up these answers to these questions,” Advani said in a 2022 podcast interview.
The waste is that the work repeats. “we’re spending a few hours every single time going to PubMed, collecting all the literature that exists,” he said of his own shifts, while a colleague down the hall had answered the same thing a week earlier. InpharmD puts the repeat rate near a third.
The pharmacist who watched Domino’s
Advani trained at Oxford, took his PharmD at the University of Georgia, and spent ten years on drug information at Atlanta Medical Center and ten years running a drug information center at Mercer. The idea came to him about a decade before 2018, WABE reported, after Domino’s launched a pizza ordering app. He built a tagged question and answer database for his own hospital first. “We actually became the number one used database at that hospital level,” he said later, ahead of the commercial reference works clinicians already had. He quit in 2018, recruited software architect Tulasee Rao Chintha as co-founder, and the two went through Y Combinator in winter 2021.
Who actually writes the answer
The 2018 model was a switchboard. About 100 academic drug information centers operate nationally, staffed mostly by pharmacy students, and InpharmD relied on four of them, Mercer included, WABE reported. A 2020 paper in Annals of Pharmacotherapy co-authored by Advani described that as a collaborative among otherwise independent academic centers, piloted at a large health care system with positive user satisfaction.
The 2026 model is not that. Answers on the company site carry a two part byline: a named pharmacist and the machine. Neil Patel, PharmD, BCPS, plus InpharmD AI. The company now trains its own, through a one year drug information fellowship paying $47,000.
The automation sits in front of the pharmacist, not after. By the company’s account, a language model decides what data is needed, a pharmacist forms a researchable question, the system generates sub questions, and those run against roughly 2 billion vector embeddings drawn from 30 million documents. The engine drafts a summary. A drug information pharmacist then approves it, declines it, or sends it back.
How much that reader has to catch is on the record, from a 2024 write up by an investor published as a sponsored post: “Right now when a query goes in, about one quarter of the literature it pulls is irrelevant and 15% is missed.” The same piece put urgent turnaround at about 46 minutes.
Selling to the chief pharmacy officer
Wellstar Health System, the largest in Georgia, signed a three year contract in July 2018 after a pilot that Advani told WABE had saved two Atlanta hospitals $5 million in a year. No method has been published for that figure. Time savings is not what closes the deal, he said later. Drug budgets are. “For every one of our systems their drug budget is 300 to 500 million every year,” he said in 2022, and about half of InpharmD’s work feeds the pharmacy and therapeutics committee that sets the formulary.
By December 2023 the company claimed 11 health systems, over 100 hospitals and 10,000 clinicians, with 18 more waiting. It raised $6.05 million led by 645 Ventures that month while calling itself already profitable.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Federal record | No 510(k), no PMA, no registered trial, no NIH or NSF award appears under InpharmD as of Sep 24, 2026. | Public record |
| Regulatory frame | FDA’s Clinical Decision Support Software guidance, docket FDA-2017-D-6569, revised January 2026, sets the section 520(o)(1)(E) criteria for software excluded from the device definition. Its FAQ states: “Software that is meant to support time-critical decision making does not meet the definition of Non-Device CDS.” InpharmD sells an urgent tier answered in about 46 minutes, and has published nothing on where it sits against those criteria. | Public record |
| Peer reviewed work | Three PubMed records name InpharmD, each with a company author: the 2020 Annals of Pharmacotherapy report, a Feb 2026 AJHP commentary on AI and the pharmacist’s role, and a Sep 2026 reply to a Touro College of Pharmacy letter about it. None studies the product. | Public record |
| Accuracy, as claimed | Qlarant Capital’s Jan 2023 release: “When compared to human controls, it has achieved a 94% accuracy rate.” No question set, sample, comparator or scoring method published. | Partner-stated |
| Accuracy, tested | The nearest published benchmark, Ipema et al., June 2026, scored OpenEvidence, Clair, GlassHealth, DougallGPT and ChatGPT against drug information pharmacists on 30 questions using the validated 25 point CLEAR framework. OpenEvidence led at 17.82 of 25; every chatbot scored statistically below the pharmacists. InpharmD was not among them. | Public record |
| Named customers | Wellstar, three year contract, July 2018, reported by three Atlanta outlets. UCLA Health, OhioHealth, Emory Healthcare and Piedmont Healthcare named in company posts, September 2026. | Independent and company |
| Customer metrics | UCLA Health: 849 inquiries, April 2025 to April 2026, an estimated 8,930 hours of pharmacist time saved. OhioHealth: one quarter at 2,872 hours, $90K, 4.6 stars. Emory pilot: 866 hours, $30,000 monthly. Piedmont: 4x ROI in Q2 2026. All company posts, none audited. | Company-stated |
| Return on investment | In 2022 Advani said the return was “like 20 X or, or even greater in some cases.” The 2026 home page advertises 4x or better across more than 30 health systems. | Claim has changed |
Read plainly: the commercial record is real and getting easier to check, because the customers now have names and numbers attached. The clinical record is not the same thing. Every accuracy figure attached to InpharmD comes from the company or an investor in it, and the only rigorous published comparisons in this category tested other tools and found all of them below the pharmacists. That argues for keeping a human in the loop, which is what InpharmD does. It is not yet evidence about this company’s output.
What to watch
- A peer reviewed evaluation of InpharmD’s own answers against pharmacist answers on a validated instrument. None exists as of September 2026.
- Whether the 94 percent accuracy claim is ever backed with a published question set and method.
- Whether a named customer publishes its own audit rather than a vendor post.
- How the company positions itself against FDA’s Non-Device CDS criteria.
- A new financing. No Form D has ever been filed under this company name.
In their words
“It could be a simple dosage question, or it could be something that I would have to look up, a complete disease state that I’ve never heard of.”
Neil Patel, drug information pharmacist, to WABE, 2018 · Independent
“20% of their days are just spent on PubMed, looking up these answers to these questions.”
Ashish Advani, CEO, Founders and Builders podcast, 2022 · Interview
“InpharmD has a extremely impressive proprietary data set and the pilot conversions are exceeding any benchmarks in the industry”
Jon Smith, VP, 645 Ventures, in the same December 2023 piece · Investor
“Our partnership with InpharmD continues to be an unqualified success.”
Scott McAuley, VP of Pharmacy and Laboratory Services, Piedmont Healthcare, in a company post, 2026 · Customer, via company
“Pharmacist expertise remains essential for ensuring answers to DI questions are of high quality.”
Ipema et al., Journal of the American College of Clinical Pharmacy, June 2026 · Peer reviewed
Related companies
Sources
- Public recordClinical Decision Support Software guidance and FAQs, FDA-2017-D-6569
- Public recordDevelopment of a Drug Information Service Collaborative in Academia
- Public recordAI and the future of the pharmacist’s role in drug information
- Public recordUnni et al. response, and the authors’ reply
- Public recordEvaluation of Clinically-Focused AI Chatbots for Drug Information Questions
- Public recordFDA, ClinicalTrials.gov, NIH, NSF, SEC EDGAR: no records
- IndependentTech Startup InpharmD, WellStar Sign Deal To Provide Mobile Medical App
- IndependentTech startup InpharmD signs contract with WellStar Health System
- InterviewAshish Advani of InPharmD on Saving Hospitals Millions of Dollars
- IndependentInpharmD Raises $6.05M in Funding
- IndependentAI and Drug Information: Where OpenEvidence Stands Out
- InvestorWhy We Invested in InpharmD
- SponsoredClinical Pharmacists, Generative AI, and InpharmD, sponsored post by an investor
- PartnerQlarant Capital Becomes Investor in InpharmD
- CompanyAnnouncing Our $6.05m Seed Round, and Decoding InpharmD
- CompanyHome page, FAQ, fellowship page and inquiry retrospective
- CompanyCustomer posts naming UCLA Health, OhioHealth, Emory and Piedmont
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
