Breaking Exactrx codes the surgery and preps the claim in under 30 seconds 96% criterion-level accuracy across 3,145 clinician-reviewed cases Athena Doshi named to Inc.'s 2025 Female Founders 500 0% clinical hallucination vs. 28-39% for standard LLMs Validated at Weill Cornell and Northwestern Medicine 80,000+ payer rules checked per case Breaking Exactrx codes the surgery and preps the claim in under 30 seconds 96% criterion-level accuracy across 3,145 clinician-reviewed cases Athena Doshi named to Inc.'s 2025 Female Founders 500 0% clinical hallucination vs. 28-39% for standard LLMs Validated at Weill Cornell and Northwestern Medicine 80,000+ payer rules checked per case
The Company Desk — Health & AI Nashville, Tennessee · Founded 2024

Exactrx reads the note. Then it bills the case.

A healthcare-AI startup that turns operative notes into submission-ready claims - and leaves the button to a human.

Exactrx logo
The Exactrx wordmark. A company that would rather show you the receipts - 3,145 cases, 96% accuracy - than tell you it's revolutionary.
96%
Criterion Accuracy
80K+
Payer Rules / Case
<30s
To Draft a Claim
~18
Employees
The Story

The quiet leak in the operating room

There is a specific kind of money that healthcare loses without anyone stealing it. The surgery happens. The care is delivered. The note is written. And then the claim goes out slightly wrong - a missing modifier, an unmet payer requirement, a documentation gap - and it comes back denied. Multiply that across a surgery center's year and you get a number that would make a CFO wince: hundreds of thousands of dollars, gone not to fraud but to friction.

Exactrx, a company founded in 2024 and run out of Nashville and New York, is built around that leak. Its pitch is unglamorous in the best way: when a surgical case wraps, the software pulls the documentation from the electronic medical record, reads the operative note, codes the case, checks every charge against more than 80,000 payer-specific rules, figures out what's missing, routes the requests to the right people, and hands back a claim that is ready to submit. All of this, the company says, before a human on the billing team touches it.

The interesting part is where the automation stops. Exactrx does not hit submit. The AI prepares; a person reviews and sends. In an industry sprinting toward full autonomy, that is a deliberate speed bump - a design choice that says the company understands exactly whose signature is on the line. As founder and CEO Athena Doshi puts it: "We prepare your claims. You review and submit." It is the rare AI product that brags about knowing where to stop.

Doshi does not describe Exactrx as billing software. She describes it as "access-to-care infrastructure," which sounds like a reframe until you follow the logic: a denied claim is not just a finance problem, it's the reason a surgery gets delayed or a facility can't afford to keep its doors open. The revenue and the access are the same lever. That framing decides which fight the company thinks it's in.


"I watched denials quietly break the financial engine."
Athena Doshi — Co-founder & CEO, Exactrx
How It Works

From operative note to ready claim

01 / READ

Pull & parse

Vision-language models pull documentation from the EMR and read the operative note the way a coder would.

02 / CODE

Code the case

The engine assigns codes and generates charges for the procedure performed.

03 / CHECK

Validate rules

A tri-engine system scores the claim against 80,000+ payer-specific rules and flags what's missing.

04 / ROUTE

Human submits

Requests route to the right people; a specialist reviews complex cases, then a human hits submit.


The Receipts

A company that publishes its numbers

Plenty of healthcare startups ask you to trust a black box. Exactrx did the opposite and ran the study. Across 3,145 cases reviewed by MDs, DOs, RNs and APPs at Weill Cornell Medicine and Northwestern Medicine, the company reports 96% criterion-level accuracy - and, more strikingly, 0% clinical-criteria hallucination, against a 28-39% range it cites for standard large language models.

Exactrx
96% accurate
Exactrx halluc.
0%
Std LLM halluc.
28-39%

Source: Exactrx validation study, 3,145 clinician-reviewed decisions. Figures reported by the company.


What You Can Do With It

Three ways it earns its keep

Precision RCM

Autonomous claims

Reads the note, codes the case, validates charges, and produces a submission-ready claim in under 30 seconds - humans review and send, specialists take the complex ones.

Analytics

Revenue intelligence

Denial tracking, payer-contract analysis, underpayment recovery, and root-cause analysis for every adjustment - so a facility can see where money leaks and why.

Pre-Op Checklists

Fewer cancellations

Embedded, AI-powered checklists surface payer requirements before surgery, cutting day-of cancellations and shaving prep time per case.


The Founders

A pharmacist, a professor, and an engineer walk into billing

Hard problems tend to live at the seams between disciplines, and the Exactrx founding table is built out of seams. Athena Doshi is a two-time digital-health founder - her prior public-health SaaS platform was acquired in 2021 after growing to serve over 1,000 organizations - with a background in physiology and neuroscience. The rest of the team fills in the parts of the problem she can't cover alone.

Athena Doshi

Co-founder & CEO

Two-time healthtech founder, ex-Omada Health and Heartbeat Health. Named to Inc.'s 2025 Female Founders 500.

Dr. Jan Krumsiek, PhD

Co-founder & Head of AI

Professor of AI and computational biology at Weill Cornell Medicine; 15+ years in biomedical AI, NLP and large language models.

Ellard Li

Co-founder & CTO

Software engineering leader with 15+ years across digital health and fintech; formerly a staff engineer at WePay.

Dr. Gina Cocos, PharmD

Co-founder

Clinician with a PharmD from UCSF and experience across ASC, hospital, and outpatient settings.


"I think about clinicians who went into medicine to help people and spend half their day on paperwork instead."
Athena Doshi — Co-founder & CEO

The Money & The Milestones

Lean, backed, and building pilots

Exactrx is early and it isn't pretending otherwise. It carries a reported seed raise of around $170,000, backing from Techstars, Tampa Bay Wave, MassChallenge, Startup:NYC and Femovate, and in 2025 opened a Wefunder community round at a reported $12 million valuation. Sales are founder-led with minimal marketing - and by late 2025 the company reported roughly 85 qualified opportunities and seven pre-pilot commitments, with pilots slated to launch in the first quarter of 2026.

  • Seed round: ~$170,000 (reported), 2024 - Techstars, Tampa Bay Wave, MassChallenge, Startup:NYC, Femovate.
  • 2025 Wefunder community round at a reported ~$12M valuation cap.
  • Fast Company World Changing Ideas 2024 honoree; named among Business Worldwide's "20 Most Innovative Companies to Watch" (2023).
  • Reported ~85 qualified opportunities and 7 pre-pilot commitments; pilots launching Q1 2026.
Latest Updates

The last twelve months

2025-06

Athena Doshi named to Inc.'s 2025 Female Founders 500 list.

2025

Launched a Wefunder community round at a reported $12M valuation; reported ~85 qualified opportunities and seven pre-pilot commitments.

2026-Q1

Pilots scheduled to launch with surgical facilities.


Worth Knowing

Five things that stick

  • Exactrx refuses to auto-submit claims on purpose - the AI prepares, a human hits send.
  • It checks each case against more than 80,000 payer-specific rules in under 30 seconds.
  • Its head of AI is a professor of AI and computational biology at Weill Cornell Medicine.
  • The company frames itself as "access-to-care infrastructure," not billing software.
  • US healthcare burns an estimated $166 billion a year on administrative overhead - the pond Exactrx is fishing in.

Watch & Read

Interviews & demos

Find Exactrx

Links & profiles

Share This Profile

Caption: The founders keep a house rule - the model never signs the claim. It drafts. A human sends. In a business built on trust with payers and patients, that last click is the whole product.