CHAPTERS
- 0:00 – 0:39
Rejections, prototypes, and the “game starts now” startup mindset
Tanay recounts being rejected from YC twice before getting accepted with a working prototype. He and Ankit discuss how quickly the market teaches focus: get in front of customers, solve one concrete problem, and realize that early traction is only the beginning.
- •YC application journey: rejected in high school, accepted on the third try with a prototype
- •Early lesson: customer proximity beats abstract planning
- •Starting small can still lead to rapid revenue growth
- •Psychological shift: hitting milestones (like revenue) doesn’t mean the journey is over
- 0:39 – 2:24
What Athelas Commure builds: an AI operating system for providers
Tanay explains the company’s product suite for provider organizations: ambient documentation, revenue cycle/payment automation, and patient engagement. The thread is reducing operational burden across both clinical and back-office workflows at scale.
- •Commure Scribe for ambient documentation
- •Athelas Revenue Cycle as “Stripe for healthcare,” including LLM-driven denial appeals
- •Patient engagement tools for reminders, prep instructions, and billing explanations
- •Scale metrics: billions in payment volume; tens of millions of appointments documented
- •Platform spans clinicians, billing teams, and CFO-level reporting
- 2:24 – 5:34
How AI fits into healthcare ops: automating payer-facing work
The discussion focuses on why LLMs are suddenly capable of handling real operational tasks—calling, negotiating, and appealing with payers. Tanay frames the provider/patient as protagonists and payers as the opposing force, with software designed to shift leverage back to providers.
- •LLMs recently became viable for human-like voice/text interactions
- •Healthcare systems employ huge teams for repetitive admin tasks
- •Offshoring RCM worked “okay,” but LLMs enable a software-first rewrite
- •Core workflow: documentation → claims → collections → financial reporting
- •Incentive misalignment: payer vs patient goals drives administrative friction
- 5:34 – 7:40
Origin story: YC inspiration, early computer vision, and a hacked smartphone microscope
Tanay describes growing up in the Bay Area with YC as a cultural influence, then building an early computer-vision prototype at YC Hacks. That experiment—classifying malarial cells—became the seed of a diagnostics-focused medical device company.
- •Early exposure to YC culture via Startup School and community heroes
- •2014-era ML constraints: segmentation + Random Forests, pre-modern deep learning
- •Prototype: cheap optics + smartphone microscope + classifier for infected cells
- •Science fair obsession becomes the company’s initial direction
- •Thesis: bring lab-grade analysis closer to patients using software + hardware
- 7:40 – 10:33
Getting into YC: co-founder strengths, Stanford resources, and early conviction
Tanay explains why they believed they could build in a regulated domain at a young age: complementary co-founder skills and Stanford mentorship. Early support from respected researchers/investors reinforced their willingness to tackle healthcare complexity.
- •Co-founder Deepika’s bioengineering/microfluidics background anchored feasibility
- •Stanford mentorship and AI lab exposure expanded ambition
- •Richard Socher’s early belief and first check post-YC
- •Applied to YC immediately, rejected twice, accepted after Stanford-based prototype
- •Commitment to a hard, regulated path rather than avoiding complexity
- 10:33 – 14:17
Clinical trials at startup speed: first-principles approach to FDA timelines
In YC, they discarded “multi-year” medical device assumptions and compressed the schedule aggressively. Tanay details how they ran a lean clinical trial by choosing a fast partner site and eliminating nonessential process overhead.
- •YC shock: med device timelines challenged as mostly process/paperwork
- •Goal set: complete clinical trial + near-ready FDA submission within the batch
- •Rejected a $120K/3-year default trial plan; rebuilt the plan from scratch
- •Executed trial in Juarez, Mexico to move quickly with a willing partner
- •Founders ran the trial themselves; compared device output vs standard Sysmex system
- 14:17 – 16:30
From FDA clearance to market reality: early commercialization and limits of a device-only business
After YC, the company focused intensely on FDA clearance and ruggedizing the device for real-world use. They found early commercial pull—including a major pharma contract—then realized the ceiling of a narrow med-device trajectory versus a broader software platform opportunity.
- •18–24 months of single-threaded focus on FDA clearance and reliability testing
- •First major customer: pharma company supporting Clozapine monitoring
- •Commercial proof came even before approval based on clinical results
- •Post-clearance growth to ~$3–4M revenue highlighted traction—and new challenges
- •Strategic realization: could be acquired as a good device outcome, or expand scope dramatically
- 16:30 – 21:54
Pivot to software (2020): expanding from 1% of patients to the whole practice
By deploying hardware in clinics, they earned trust and uncovered adjacent operational pain points—manual uploads, portals, faxes, claims steps. COVID accelerated demand for remote care, and the team moved into telehealth, monitoring, and early payments/claims automation.
- •Clinic immersion surfaced workflow breakdowns that weren’t visible from the outside
- •Trust earned via hardware deployments enabled deeper problem discovery
- •COVID drove urgent need for remote monitoring and telehealth tooling
- •Started with narrow workflow/claim automation for existing patient cohorts, then expanded
- •Strategic framing: increase “share of wallet” by serving the full patient panel
- 21:54 – 27:08
Physician burnout and the shift from CIO-led to doctor-led software adoption
Tanay argues prior digitization (EHRs) primarily served compliance and billing, not clinicians, increasing the “work tax.” With LLMs and post-COVID stress, physicians are now the primary champions of tools that directly save them time.
- •EHR-era software optimized for CFO/CIO priorities: payment and legal defensibility
- •Administrative bloat degraded physician productivity and job satisfaction
- •Consolidation pressures pushed doctors into large systems for overhead coverage
- •LLMs create a “generational opportunity” to remove documentation and back-office burden
- •Key change: physicians are now the loudest internal advocates for adoption
- 27:08 – 30:16
Ambient documentation at scale: self-serve growth, HIPAA realities, and enterprise expansion
Commure Scribe is framed as one of the fastest-moving LLM adoption categories in healthcare. Tanay describes how self-serve physician signups (even via personal credit cards) create bottom-up momentum, which then expands into enterprise integrations.
- •Ambient scribing: listen → summarize → generate documentation for billing workflows
- •Growth: from ~100K appointments (2023) to ~20–25M/year; self-serve reaching ~5M
- •New behavior in healthcare: doctors discover, buy, and adopt tools directly
- •HIPAA compliance can be achieved; biggest friction is hospital internal policy, not law
- •Technical scaling challenges: unreliable networks, offline mode, background uploads
- 30:16 – 32:43
“Stripe for healthcare”: rebuilding revenue cycle with models, monitoring, and denial defense
Revenue cycle is presented as a high-stakes, adversarial interface with payers—closer to cybersecurity than billing. The system must detect failures, incomplete payer APIs, and denial tactics, then respond automatically to protect physician cash flow.
- •RCM scope: collections, claim submission, denials, appeals, BI/finance visibility
- •Payer friction patterns: lost claims, delayed denials, timely filing traps
- •Payer APIs work inconsistently; handling edge cases at scale is the core challenge
- •Models monitor claims status and orchestrate responses (including appeals/negotiation)
- •Word-of-mouth flywheel: “my revenue went up by 15%” drives rapid adoption
- 32:43 – 36:28
Platform strategy and distribution: bundling, point solutions, and why single-feature companies plateau
Tanay explains why healthcare winners tend to become platform companies: complexity, cash flow dynamics, and distribution constraints. He argues many point-solution LLM startups will stall, while platforms can bundle and expand efficiently.
- •Thesis: $100B healthcare outcomes require platforms, but you must start as a point solution
- •Distribution is the hardest problem; platform + RCM backbone reduces go-to-market friction
- •Doctor-led adoption enables new wedges, but point solutions still hit scaling walls
- •Bundling parallels Microsoft: feature becomes native and wipes out standalone businesses
- •Prediction: many ambient scribe wrappers will disappear when Epic turns on equivalents
- 36:28 – 38:45
Competing with Epic: speed, systems of engagement, and scaling through forward-deployed teams
Tanay positions Commure as the fast-moving layer built around/above core EMRs, especially for health systems that avoid Epic’s cost and roadmap dependency. The company relies on physician-driven value, plus forward-deployed engineering pods that co-develop solutions onsite.
- •Epic’s counter-argument mirrors Commure’s—competition is about platform leverage
- •Speed as a core value: some systems won’t wait years for Epic roadmaps
- •HCA and Tenet cited as highly profitable systems that avoid Epic implementations
- •Commure targets the “system of engagement” on top of EMRs
- •Forward-deployed engineering teams work directly with medical directors and clinicians to create new products
- 38:45 – 47:34
The next decade of medicine: telehealth, home sensors, and AI copilots reshaping care delivery
They zoom out to patient experience and where change is most likely: more care shifting to home, better no-show reduction via engagement, and doctors reclaiming time through LLMs. Tanay forecasts a split between high-acuity in-person care and low-acuity self-service/virtual care, with AI copilots accelerating diagnosis and decision-making.
- •Patient experience hasn’t changed much, but telehealth and automation are growing on the fringes
- •Engagement example: AI-guided colonoscopy prep improves no-show rates materially
- •LLMs save physicians hours/day, potentially expanding specialist availability
- •AI copilots may outperform humans in assessment and care planning, with clinicians as responsible sign-off
- •Forecast: ubiquitous sensors/wearables; insurance becomes more catastrophic; routine care shifts out-of-pocket and out of payer complexity
- 47:34 – 49:33
Advice to founders: start tiny, live with customers, and learn distribution by doing
Tanay closes with a playbook for healthcare founders: don’t fear small initial markets, and prioritize customer-embedded problem solving. He also suggests that working inside scaling companies can be a fast path to learning distribution and operational execution in complex industries.
- •Get in front of customers; observe nuances that desk research can’t reveal
- •Start with one narrow problem—even if the initial TAM looks “laughable”
- •Earn trust through delivery, then expand scope based on real pull
- •Distribution lessons matter as much as product in healthcare
- •Career advice: roles like forward-deployed engineering can build founder skillsets quickly
