For hotel groups & hospitality operators

Your profit is in the banquet hall. Your risk is in the guest file.

Hotel software has spent a decade optimizing the room. Meanwhile the banquets, restaurants, and spa that carry the margin run on spreadsheets, and the guest data that carries the regulatory risk runs on nothing at all. Vaiyu builds the spine under both: one guest identity, one revenue truth, and models your revenue manager can operate and your auditor can reconstruct. We have been doing exactly this in hospitals for a decade.

Start a conversation See the track record

15+

years operationalizing AI, prototype to production

$9M+¹

federally funded AI R&D led

up to50%²

training cost cut for clients

71³

sites in one federated learning study

up to90%⁴

inference latency removed

The burning platform

Every system you run was designed to sell a room.

Hospitality is the most vendor-saturated software category there is, and none of those vendors is paid to solve any of the six problems below. Each one is a data, modeling, or governance problem, sitting underneath the applications rather than inside them.

01
Your systems don’t talk, and no vendor is paid to make them.
Only 11% of hotels report a fully integrated tech stack. 27% run seven or more platforms, 27% lose eleven or more hours a week to manual reconciliation, and 91% still rely on manual reporting. Every vendor sells a system of record and profits from owning its slice; harmonizing across them is nobody’s product.^a
02
The margin is where the software isn’t.
Non-room revenue (banquets, weddings, restaurants, spa) reaches up to 60% of the total in Indian Tier II/III hotels, against roughly 32% for US full-service properties. India runs eight to ten million weddings a year, in a market estimated near US$130bn. Every revenue-management system on the market prices rooms.^b,c
03
Your room rate is now a tax decision.
Since 1 April 2025, a single room sold above ₹7,500 on a single night makes a property “specified premises” for the whole of the following financial year, reclassifying every restaurant inside it. The declaration is filed a year ahead, under demand uncertainty, and continues until you revoke it. No pricing system models this, because none of them know what a specified premises is.^d
04
Agents book what they can read.
AI Overviews cut click-through to the top organic result by roughly 58%, and organic click-through by 61% on affected queries. Travel is the fastest-growing vertical for AI referral traffic. ChatGPT launched apps with two of the largest OTAs; Google published an agent-payments protocol with sixty-plus backers. Inventory that is machine-readable only inside an OTA is inventory the OTA owns.^e,f,g
05
The commission is not the whole cost.
OTA commission runs 15–25%, and to about 30% with promoted placement, against 5–12% to acquire a guest directly. OTA bookings also cancel harder: one major OTA’s group segment at 42%, against 11.1% for direct. Independents take 61% of bookings through OTAs; branded chains, 22%.^h
06
A dated regulator is coming for the guest file.
India’s DPDP obligations commence 13 May 2027, with penalties reaching ₹250 crore for a security-safeguards failure. A hotel group holds passport scans, payment credentials, and (through a spa) treatment records that are special-category health data for inbound EU guests. The Aadhaar photocopy at your front desk is already outside what UIDAI permits.^i,j,k

What transfers

We didn’t learn this in hospitality. That’s the point.

We have spent a decade building AI that runs in hospitals, under regulators who ask for evidence rather than a demo. Nothing below is a stretch: it is the same work, applied to a guest instead of a patient.

Harmonization across scanners, sites, formats, and legacy systemsOne guest identity and one revenue truth across PMS, POS, booking engine, banquet, spa, and bakery.
Validation protocols and audit-ready documentation for clinical reviewData-fiduciary obligations met on a dated plan, and a pricing decision an auditor can reconstruct a year later.
Privacy-preserving design: de-identification and access controlPassport, Aadhaar, payment, and spa-treatment records, held lawfully and consented at every touchpoint.
Inference optimization for constrained clinical hardware⁴Offline-capable check-in and kitchen tooling at a resort with no reliable uplink.
Low-code frameworks so non-ML staff can run the modelA revenue manager, not a data scientist, operates the banquet model.
Reproducible pipelines and experiment lineageWhy the model quoted that rate, for that wedding (a year later, in a dispute).

Your spa is a clinic in the eyes of a data regulator. Clinics are where we come from.

What we do for hotel groups

Six ways in: your systems, your data, your team afterward.

We do not build or replace a PMS, a booking engine, a channel manager, or a point-of-sale. Those work, you already pay for them, and ripping one out is a multi-year project with its own vendor category. We integrate with what you run, and build the layer no vendor is incentivized to build.

Readiness & AI strategy sprint

Two to four weeks, fixed scope and fee: use-case triage against the P&L, a data-readiness audit, and an explicit build-vs-buy verdict per use case, including, wherever it is true, “just buy the software.”

The guest & revenue spine

Harmonization across PMS, POS, booking engine, banquet, spa, and the spreadsheets. One guest identity, one revenue truth per property per day, one consent ledger. Boring, load-bearing, and everything above it depends on it.

Banquet, F&B & demand models

The models nobody sells: event conversion priced against the room inventory an event displaces, multi-outlet forecasting and kitchen-waste reduction, tax-aware rate strategy, and a review-response model that knows which reviews are worth answering.^l,m

Guest technology & deployment

Lawful digital check-in, automated foreign-guest reporting inside the statutory 24 hours, tokenized card handling, agent-readable distribution, and offline-capable inference for properties without a reliable uplink.^n,o

Guest-data governance

A dated plan to 13 May 2027: consent architecture, retention and erasure that actually delete, breach response, model audit, and identity capture at the front desk that is lawful today, not next year.^i,k

Fractional AI leadership

Senior data and AI leadership for a group that will never hire a full-time Chief Data Officer, and shouldn’t. Vendor vetting, roadmap governance, and board reporting, a few days a quarter.

Credibility

Peer-reviewed, cited, and standard-setting.

Our team’s research appears in Nature Communicationsand Nature Machine Intelligence, and has been covered by The Wall Street Journal. We’ve led more than$9M in NIH/NCI-funded R&D, and we hold the Vice Chair for Algorithmic Development of theMLCommons Medical Working Group, helping write the standards medical AI is measured against.

Publications & standards work →

Vice Chair, Algorithmic Development, MLCommons Medical Working Group
Organizers of leading benchmarks: the BraTS and FeTS challenges (MICCAI)
Contributors to imaging standards: IBSI and AI-RANO
Reviewers for Nature Communications, IEEE Transactions on Medical Imaging, and Radiology
Tutorial faculty at MICCAI, AAAI, ISBI, and RSNA

Built in the open

These frameworks were built so hospitals could train together without ever sharing a patient record. The same harmonization, privacy, and lineage machinery is what a hotel group needs to hold a guest file in 2027.

Plus 40+ conda-forge packages maintained for reproducible scientific computing.

How we engage

Start at the front desk. Finish at the P&L.

Each of these is a standalone engagement with its own scope, fee, and exit criteria. Most groups start with one of the first two.

Readiness sprint

Two to four weeks, fixed scope and fee: where AI pays, where it doesn’t, and what your data can actually support today.

A costed roadmap your owner and your GM both believe.

Check-in pilot

One property. Lawful digital check-in, automated foreign-guest reporting, tokenized cards, live, and measured against the front desk it replaces.

A visible win, and the guest-data spine underneath it.

Banquet & F&B model

The event funnel and the kitchens, forecast and priced at one property, with a before-and-after on conversion, covers, and waste.

Margin found in the half of the business nobody modeled.

Tax-aware pricing review

Annual. We model the specified-premises decision per property before the filing window closes; your chartered accountant files it.

A filed position with the sensitivity analysis behind it.

Data-protection readiness

A dated program to 13 May 2027: consent, retention, erasure, breach response, and lawful identity capture at the desk.

An obligation met before it becomes a penalty.

Agent-readable distribution

Structured supply, machine-readable rates and availability, and an evaluation harness for when an agent gets a booking wrong.

Bookable by an AI agent that isn’t an OTA.

We model the specified-premises decision; we do not file it. Any tax position we recommend goes to your chartered accountant for sign-off, in writing, before it is acted on.

The room is a commodity. Everything else is yours.

Start with a readiness sprint or a single-property check-in pilot. If banquet intelligence, agent-readable distribution, or guest-data readiness is on your roadmap, we should talk.

Write to ussupport [at] vaiyu [dot] solutions

Sources & attribution

1. Led by our founder across NIH/NCI-funded programs at the University of Pennsylvania and Indiana University.
2. Vaiyu client engagements: pre-training optimization with model accuracy maintained or improved.
3. Pati, S. et al. “Federated learning enables big data for rare cancer boundary detection.” Nature Communications 13 (2022).doi:10.1038/s41467-022-33407-5; 71 sites across 6 continents, the largest real-world federated learning study to date.
4. Founder track record at Indiana University: inference latency reduced by up to 70%, compute requirements by 10–50%, in clinical research environments.
a. Otelier, 2026 Hotel Operations Index. A vendor-sponsored survey; the direction is corroborated by Cloudbeds’ 2026 State of Independent Hotels (67% of independents name managing disparate systems a top concern) and by NYU SPS Tisch / RateGain / HEDNA,State of Distribution 2025 (80% of hotels spend up to two days a week on manual reporting; ~21,000 properties).
b. Savills India with Hotelivate, survey of ~600 hotels across 150 cities, 2024–25: non-room revenue up to 60% of total in Tier II/III. US contrast (~32% non-room): CBRE Hotels Research,Trends®, 2025.
c. Jefferies with CAIT, 2023: India’s wedding market estimated near US$130bn across 8–10 million weddings a year; the second-largest consumption category after food and grocery.
d. CBIC FAQ on restaurant service supplied at “specified premises,” 27 March 2025, and Annexure VII: a premises is specified if, in the preceding financial year, the value of supply of any unit of accommodation exceeded ₹7,500 per unit per day; restaurant service there attracts 18% GST with input tax credit, elsewhere 5% without. Effective 1 April 2025; the definition and its preceding-year lookback survive the rate rationalization effective 22 September 2025. Opt-in and opt-out are filed 1 January – 31 March for the succeeding financial year and auto-continue until revoked (GSTN Advisory No. 645, 4 January 2026). Verified 9 July 2026. This is a modeling problem, not tax advice.
e. Ahrefs, December 2025, ~300,000 keywords: AI Overviews cut click-through to the top-ranking page by ~58%. Seer Interactive, September 2025: organic click-through 1.76% → 0.61% on queries showing an AI Overview (3,119 queries; 25.1M impressions). Both are general-query studies; no hospitality-specific measurement exists.
f. Similarweb, 2025: travel and hospitality is the fastest-growing vertical for AI referral traffic (+1,700%, July 2024 – February 2025).
g. OpenAI DevDay, 6 October 2025: ChatGPT apps launched with two of the largest OTAs as partners. Google Agent Payments Protocol, 16 September 2025, 60+ partners. Note that OpenAI walked back in-chat travel checkout in March 2026; transactions remain on the OTA’s own site. The near-term shift is in discovery, not payment.
h. Commission ranges are directional; no OTA publishes a rate card. The firmest anchor is Booking Holdings’ FY2024 blended take rate of ~14.3% (revenue ÷ gross bookings). Cancellation by channel: D-EDGE, October 2024: direct 11.1%, Booking.com group 42% (Europe). Channel mix: Cloudbeds, 2024 (independents 61% via OTA) and RateGain, 2025 (branded chains 22%).
i. India’s Digital Personal Data Protection Rules, notified 13 November 2025. Substantive obligations (notice, security safeguards, breach notification, retention, erasure) commence 13 May 2027. Penalties reach ₹250 crore for a security-safeguards failure (DPDP Act 2023, Schedule 1). Consent notices are required in all 22 scheduled languages.
j. GDPR Article 9: health data is special-category, and applies to EU-resident guests’ spa and wellness treatment records. India’s DPDP framework draws no equivalent sensitive-data class.
k. UIDAI / PIB advisory, May 2022: private entities such as hotels may not collect or store Aadhaar photocopies. Masked Aadhaar, offline e-KYC, QR, or OTP verification is the permitted route.
l. Champions 12.3 (WRI/WRAP-backed), The Business Case for Reducing Food Loss and Waste: Hotels, 2018: 21% average kitchen food-waste reduction in year one across 42 hotel sites in 15 countries; $7 saved per $1 invested. The strongest non-vendor evidence in the category.
m. Anderson, C., “The Impact of Social Media on Lodging Performance,” Cornell Center for Hospitality Research, 2012: a 1-point gain on a 100-point review index is associated with +1.42% RevPAR. Anderson, C. & Han, S., Cornell CHR, 2016: revenue rises with review-response rate only to ~40%; responding to more than ~85% of reviews is associated with lower revenue than not responding at all. Which is why we model which reviews to answer.
n. Immigration and Foreigners Act, 2025 (Act No. 13 of 2025), in force 1 September 2025: every accommodation provider must report each foreign guest via the Bureau of Immigration portal within 24 hours of arrival.
o. RBI card-on-file tokenization, in force 1 October 2022: merchants and aggregators may not store the card number, expiry, or CVV; network tokens are required. Directly affects hotels holding cards to guarantee a no-show.