OMNIS

Always-on precision health.

Your body, continuously decoded.

An always-on intelligence layer that turns episodic care into continuous, MD-signed precision health across genome, biomarkers, wearables, and behavior.

scroll
01

The proposition

Omnis is the audit-grade AI clinical decision layer that sits underneath every prescribing decision a regulated-markets healthcare-services company makes — and turns episodic care into continuous, MD-signed precision health.

One stack. Sold underneath every clinic on the network. The platform compounds with multi-tenant data, the operating arm proves it works, and the regulators want it tomorrow.

01 · THE PROBLEM

Prescribing is bottlenecked on bandwidth and compliance.

Specialty pharma platforms, healthcare-services rollups, and clinic networks all face the same wall: clinical decision quality scales with clinician hours, and compliance overhead caps how fast any AI tool can get deployed.

The AI tools that would help are stuck behind BAA reviews, cross-border data-residency carve-outs, and per-token billing that scales the wrong way.

02 · WHO IT'S FOR

Buyers who already own the patient relationship.

Not a consumer brand. We sell the substrate that lives underneath these four buyer shapes:

  • Specialty-pharma platforms needing a prescribing + RWE (Real-World Evidence — the patient-outcomes data drug companies need for label expansions and post-market surveillance) layer for their pipeline
  • Healthcare-services rollups needing operational substrate that scales clinical quality faster than headcount
  • Diagnostic distribution networks needing an AI overlay on the lab volumes they already move
  • Telehealth + clinic networks needing audit-grade compliance by construction, not by policy review
03 · WHAT WE SHIP

Four pieces. One platform. Audit-grade by construction.

The platform layer every clinic on the network shares:

  • 8-gate prescribing pipeline — every protocol audited through eight explicit checks with full evidence provenance
  • Watcher trio — guideline + trial + literature monitors that re-evaluate every patient when anything changes
  • On-device clinical AI — Gemma 4 31B model runs locally. PHI (Protected Health Information) never leaves the building, so HIPAA (US health-privacy law) / PDPA (Singapore + Thailand) / DPDP (India) compliance becomes a wiring diagram instead of a policy review
  • Frontier-compound readiness — the network is positioned 6–12 months early on the upcoming GLP-1+ pipeline (the Ozempic / Wegovy / Zepbound drug class, plus the next generation behind them)
04 · WHY NOW

The architecture was impossible six months ago.

Above 31 billion parameters, open clinical models stopped hallucinating — Bo Wang's UHN AI Hub benchmark, May 2026, Gemma 4 31B at 86.5% diagnostic accuracy with zero hallucinations on diagnostic tasks.

That single benchmark is what makes on-device clinical AI viable today. A year ago it was a research bet. A year from now it's a competitive cost. We're inside the window where moving first matters.

What makes it defensible

Consumer telehealth captures one slice of the patient. Single-clinic concierge captures one location. Pharma reps capture one product line. Nobody has built the audit-grade clinical decision layer underneath all of them. That layer is what we sell.

Every clinic on the network sharpens it. The competitor with no installed base cannot catch up — they have no decision history to learn from.

02

The opportunity

The market has a clear shape — every player owns a slice. None owns the layer underneath. That's the gap. And the combination of an AI-infra team with healthcare commercial reach is the right hand to fill it — at the moment the models cross the threshold that makes the architecture work.
01 · THE GAP

Every player owns one slice. No one owns the layer underneath.

Consumer brand + DTC fulfilment

Hims & Hers · Ro · Eucalyptus

Annual biomarker labs

Function · Protocol Health · MORROW

Coaching app + wearable

Healthify · Ultrahuman · Twin Health

Pharmacy distribution

Truemeds · 1mg · Apollo 24/7

Single-location clinic concierge

Biopeak · Sukino · SG saturated set

Gray-channel peptide aesthetic

ARQ · TAC · Mumbai chains

The audit-grade clinical decision + monitoring + dispensing platform underneath all of them — is unbuilt.

Pricing pressure on each slice keeps competitors inside their slice. Building the cross-slice substrate requires regulatory chops + AI infra + capital patience to sit beneath the layer that earns. We're the team where those three meet.

02 · OUR RIGHT TO PLAY

Regulated-industry AI infra × the moment the models work.

TECHNICAL TRACK RECORD

On-prem production AI at regulated institutions.

Our technical partnership has shipped on-premise AI deployments at the BSE (Bombay Stock Exchange) / NSE (National Stock Exchange) / top-15 Indian banks — the same compliance-by-construction architecture Omnis uses for HIPAA (US health-privacy law) / DPDP (India's 2023 data-protection act) / PDPA (Singapore + Thailand). Not a research lab. A production line.

The narrative arc

The gap is structural. The technical window opened six months ago (Bo Wang UHN benchmark, Gemma 4 31B). The shape of buyer who needs this layer is already operating but missing the AI overlaythat lets their clinical quality scale faster than headcount. We're the layer they all need next.

The window is wide enough to walk through. It is not wide enough to wait through.

03

Two layers, one stack

We're building two things at once. Omnis is the platform — the audit-grade clinical intelligence layer that decides every protocol, monitors every patient, logs every decision. Sold underneath every clinic on the network. Omnis Digital Clinic is the operating arm — a two-operator team running a flagship clinical front-end that proves the platform out in the markets where we deploy first.
HOW THEY CONNECT

Omnis Digital Clinic runs the front-end — patients arrive, intake runs, the gate pipeline pre-drafts a protocol, the specialist MD signs off, the protocol dispenses through a partner pharmacy, the watcher infrastructure monitors continuously. Every one of those events writes to the Omnis platform. Meanwhile a hospital network in another geography licenses Omnis underneath their existing clinical operations — zero patients shared, multi-tenant data isolated, but the gate-pipeline weights and outcome benchmarks they get back are richer because the clinic is feeding them too.

Two layers. Same engineering team. One investable thesis.

04

What the platform actually does

Every patient on Omnis moves through five stages. Each stage is a distinct piece of the platform — and each is detailed in its own section below. This is the overview.
01
Stage

Intake — biomarker baseline + life context

Every patient starts with a structured intake: bloods (HbA1c [3-month average blood sugar], ApoB [a cardiovascular-risk lipoprotein], ALT [liver enzyme], eGFR [kidney filtration], TSH [thyroid hormone], vitamin D — extendable per archetype), wearable sync (Whoop / Oura / Apple Health), full medication and supplement list, family history, current life context.

Plus archetype routing: the patient profile triages into one of five operating archetypes — Founder Optimizer, Aesthetic Mid-Lifer, Perimenopausal Reset, Metabolic Recovery Returner, Diaspora Returner. Archetype determines content layer, monitoring cadence, and acquisition channel.

02
Stage

AI protocol draft — through the 8 gates

The AI drafts a protocol against the patient's full record. Every line item — drug class, dose, titration schedule, adjunct supplements, monitoring cadence — passes through eight explicit gates: safety, efficacy, adherence, cost, drug interactions, timing, contraindications, dosing precision.

Each gate logs the evidence it consulted (which guideline edition, which lab values + timestamps, which interaction database). The output isn't just a protocol — it's a protocol plus a reproducible reasoning chain.

03
Stage

MD review + sign-off — clinician in the loop

The drafted protocol queues to a specialist physician (endocrinologist / internal-medicine specialist / cardiologist per DCGI [Drugs Controller General of India] 24 March 2026 advisory for GLP-1 [the Ozempic / Wegovy drug class]; cardiology / metabolic / aesthetic specialty per indication). The MD (medical doctor) sees the AI's reasoning, the gate-by-gate evidence, and can approve, edit, or reject.

Every MD edit becomes training signal — the rejection reason, the dose adjustment, the alternative compound preference all flow back into the gate-pipeline weights. The platform sharpens with every clinician interaction.

04
Stage

Fulfilment — dispense through partner pharmacy

Signed protocol routes to a Form-20 partnered pharmacy in the patient's revenue district (or international shipping for cross-jurisdiction patients). CDSCO (Central Drugs Standard Control Organisation — India's national drug regulator) approved compounds only — branded generics, approved biologics, no gray-market peptides. Cold chain managed. Patient receives + onboards.

Revenue model: rev share on fulfilled product flows back to Omnis. Omnis Digital Clinic handles fulfilment in-house at our operated sites; licensed partner clinics retain their existing pharmacy partnerships.

05
Stage

Continuous monitoring — watcher trio + on-device AI

Patient enters the standing monitoring loop. Guideline watcher (CPIC [pharmacogenomic dosing], ADA, FDA labels, EMA SmPC [the EU drug-label format]) re-runs gates when relevant guidelines update. Trial watcher (ClinicalTrials.gov, CTRI [India's trial registry]) surfaces enrollment matches. Literature watcher (the major medical journals — NEJM, Lancet, plus preprints on bioRxiv) flags new evidence affecting current stack.

All of this runs locally on the clinic's on-device appliance — Gemma 4 31B or Qwen3.5-35B, 86.5% diagnostic accuracy per Bo Wang's May 2026 UHN benchmark. PHI never crosses the building's edge. HIPAA / PDPA / DPDP compliance by architecture.

The through-line

One canonical patient record. Five stages of decisions, each logged with full evidence provenance. Standing watcher infrastructure that re-evaluates every record when a guideline updates, a trial opens, or new literature lands. On-device clinical AI that keeps PHI inside the building. The architectural shape is audit-grade decisions with continuous improvement. The clinical scope — obesity, metabolic, longevity, peptide-adjacent, oncology adjuncts — is configurable per partner clinic.

05

The 8 gates

Every protocol decision runs through eight explicit checks. Each gate consults specific evidence, logs its inputs, and produces a decision with confidence. Two patients with identical labs can receive different protocols because different gates flag for them — and every decision is reproducible from the logged evidence chain. Audit-grade compliance by construction.
THE MOAT, CONCRETE

Same labs. Different protocols.

This is the moat made concrete. Two patients walk in with the same biomarkers. The gate pipeline reads their full context — adherence history, insurance, family history, current meds — and routes them to different protocols. The clinician sees both reasoning chains. Audit-grade by default.

PATIENT A
Straightforward path
Labs
HbA1c 6.4% · BMI 31.4 · eGFR 76 · ApoB 105 · ALT 38
Adherence history
92% Rx fill, trailing 12 months
Insurance
Covered, no prior auth
Family history
None coded
Current meds
Atorvastatin 20 mg
Gate output
Semaglutide 0.25 mg starter, weekly. Week-9 titration target 1.0 mg.
MD review
12 minutes — signed
Provenance log
4 KB JSON · every input timestamped
PATIENT B
Three gates flag
Labs
HbA1c 6.4% · BMI 31.4 · eGFR 76 · ApoB 105 · ALT 38
Adherence history
58% Rx fill, trailing 12 months. Statin DC at week 14.
Insurance
High deductible, $400/mo OOP cap
Family history
Father — MEN-2 syndrome
Current meds
Atorvastatin 40 mg + sertraline 50 mg
Gate output
Routed away from injectable GLP-1. Rybelsus oral 7 mg starter + MD review of MEN-2 family-history detail before proceeding.
MD review
22 minutes — more depth required
Provenance log
6 KB JSON · includes MEN-2 family-history reasoning chain

Same input labs, same recommended candidate class. Gate 03 (adherence), Gate 04 (cost), and Gate 07 (relative contraindication) flagged differently for Patient B — and every flag is reproducible from the same provenance log the auditor reads.

06

The watcher trio

Imagine a medical librarian whose job is to read every new clinical guideline, every new trial result, every new dose-response paper — and tell each patient's clinician the same week if it affects their protocol. Now run that 24/7 across millions of decisions. That's the watcher trio. Standing background infrastructure on every patient record.
Why this compounds

No clinic builds this. They can't — it requires standing engineering infrastructure running 24/7 across thousands of feeds. They subscribe to it. The watcher trio is the kind of overhead that only makes sense at platform scale, sold across many clinics. The clinic that licenses Omnis is the only clinic in their geography whose patients are getting guideline-current care without a 24-month lag.

07

The HIPAA flip

The strongest technical piece of the platform — and one that was structurally impossible six months ago. Above 31B parameters, open clinical models stopped hallucinating (Bo Wang, UHN AI Hub, May 2026). That single benchmark is what makes the architecture below viable today and would have made it a liability a year ago. HIPAA (US health-privacy law) / PDPA (Singapore + Thailand) / DPDP (India) / GDPR (Europe) compliance becomes a wiring diagram, not a policy document.PHI (Protected Health Information) never leaves the clinic. Every clinic that froze AI adoption because "we're not letting our patients' data go to OpenAI" — they say yes to this. Speed wins.
Today · the standard cloud-AI architecture

Every AI feature is a BAA negotiation.

  • ·Hospital signs BAA (Business Associate Agreement — the compliance contract a hospital signs with every cloud vendor) with OpenAI / Anthropic / Google / AWS per vendor
  • ·Patient data travels to vendor servers · processed remotely · returned
  • ·Compliance officer reviews each BAA annually · vendor pricing changes pass through
  • ·Vendor downtime = clinic downtime · per-token costs scale with patient volume
  • ·Every new AI feature = new vendor BAA review cycle
With Omnis · on-device clinical AI

PHI never leaves the building.

  • ·1U appliance ships to clinic · Gemma 4 31B (Google open-weight) or Qwen3.5-35B fine-tuned · single H100 GPU · ~$30K capex
  • ·All clinical reasoning runs locally · PHI never traverses the clinic's edge
  • ·Cloud handles only non-PHI work — literature monitoring, public guideline ingestion, patient-facing UI without identifiable data
  • ·HIPAA / PDPA / DPDP / GDPR compliance becomes architectural — no vendor BAA needed for the clinical-reasoning path
  • ·No per-token billing on the clinical path · no vendor lock-in · clinic owns the appliance
The benchmark that made this real · May 2026

Above 31 billion parameters, open clinical-AI models stopped hallucinating.

Bo Wang's UHN AI Hub benchmark (May 2026) tested 188 models on general disease diagnosis. Gemma 4 31B hit 86.5% accuracy. Fine-tuned Qwen3.5-35B hit ~88%. Critically: above 31B parameters, hallucination rate dropped to zero on diagnostic tasks. The errors that remain look like the kind a careful clinician makes on a hard case — not the kind that destroys trust.

86.5%
Gemma 4 31B accuracy
General disease diagnosis benchmark
~88%
Qwen3.5-35B fine-tuned
Above the 31B threshold
0
Hallucinations · above 31B
Errors look like clinician's hard-case errors

Source · Wang et al., UHN AI Hub, May 2026 · 188-model benchmark

Why this is structural

Any specialty-pharma platform, diagnostics distributor, or generics co selling into regulated markets fails or succeeds on the same data-residency + audit-trail question. On-device clinical AI sidesteps the entire category of failure modes.The clinic's compliance officer signs off in week one instead of month six. The hospital network's CTO stops blocking the AI rollout. The pharma partner stops asking "where does the patient data live?" because the answer is "the clinic, always."

08

Frontier compounds — early readiness

A standing monitor across FDA AdComm calendars, EMA CHMP meetings, ADA / EASD / AHA / ASCO readouts, ClinicalTrials.gov Phase-3 publications, and FDA Form-483 / Warning Letter signals. When a frontier approved compound clears its next regulatory gate, the network knows the same week. Clinics on Omnis are positioned to onboard patients 6–12 months ahead of the average rollout.
The watch loop

Six monitors, one alert path

Six monitors run continuously: FDA AdComm calendars, EMA CHMP meeting schedules, ClinicalTrials.gov Phase-3 update feeds, ADA / EASD / AHA / ASCO scientific-session abstracts, FDA Form-483 and Warning Letter feeds, and Indian CDSCO Subject Expert Committee (SEC) recommendation PDFs. When any monitor fires, the relevant patient cohort is auto-segmented for the clinic team.

FDA AdComm
EMA CHMP
CT.gov
ADA/EASD
FDA 483
CDSCO SEC
09

Patient archetypes

Five archetypes carry the early product. They are not market segments — they are the operating units for protocol design, demand generation, and clinic acquisition. Each has a distinct trigger, stack, willingness-to-pay band, and acquisition channel. Click any archetype to see the depth.
10

Business model — and the unit economics

Two-layer architecture. Omnis is the platform IP — sold B2B to wellness clinic networks, hospital systems, and telehealth platforms. Omnis Digital Clinic operates flagship clinical front-ends in select markets, feeding the platform underneath. Revenue compounds across both. Numbers below are illustrative. The inputs are designed to invite scrutiny — every assumption is labelled.
BLOCK 01 · REVENUE STREAMS

Four streams, two layers

Streams A–C earn at the platform layer (Omnis). Stream D earns at the operator layer (Omnis Digital Clinic). The four compound: more clinics → more protocols → more fulfilment → more multi-tenant data → better platform AI → higher LTV.

BLOCK 02 · COST STRUCTURE

Five cost lines, mostly variable

Compute and MD minutes scale with protocol volume. Watcher operations and customer success are fixed up to a clinic threshold. Click any line for the build-up.

BLOCK 03 · UNIT ECONOMICS

Per-patient. Per-clinic. Side by side.

The visual centerpiece. Read this like a spreadsheet, not a brochure. Left is Omnis Digital Clinic (operator-shape); right is Omnis (platform-shape). The two compound — see the bottom-line note.

CLINIC DIRECT · OPERATOR-SHAPE

Per active patient · per year · Active tier

Revenue
Subscription$2,400
Protocol fees$400
Product margin$300
TOTAL REV$3,100
Costs
Compute$60
MD minutes$240
Data + infra$48
Fulfilment ops$400
Customer success amort.$150
TOTAL COGS$898
GROSS PROFIT$2,202
GROSS MARGIN71%

Services-shape margins. Comparable to regional hospital-services operators at maturity.

OMNIS PLATFORM · SAAS-SHAPE (B2B)

Per partner clinic · per year · 200 active patients

Revenue
Base license$24,000
Per-active-patient module ($60 × 200)$144,000
Per-protocol fees (800 × $5)$4,000
Rev share on fulfilment (~$30 × 200)$6,000
TOTAL REV$178,000
Costs
Compute (cloud + on-device amort.)$12,000
Clinical content + watcher allocation$24,000
Customer success amort.$14,000
Data + infra$9,000
TOTAL COGS$59,000
GROSS PROFIT$119,000
GROSS MARGIN67%

SaaS-shape. Comparable to Veeva Vault unit economics at maturity, currency-adjusted for emerging-markets clinic pricing.

Two layers, same patient. Omnis Digital Clinic earns the services margin on the people the clinic serves directly; the Omnis platform earns the platform margin on every protocol routed through any partner clinic on the network. They compound — the more clinics on Omnis, the larger the multi-tenant dataset, the faster gate-pipeline AI improves, the higher the platform LTV.

BLOCK 04 · COMPARABLE BENCHMARKS

Where the valuation conversation lives

Six comparables anchoring the multiple-band. Pure-DTC sets the ceiling for the operator layer; Veeva sets the ceiling for the platform layer. Indian listed CRDMOs + PE outcomes anchor the exit-path conversation.

11

The landscape — and the white-space

The category is crowded. Singapore is saturated. India is split between consumer-tech, peptide aesthetics gray-market, and incumbent online pharmacies. The US shows where it ends up — vertically-integrated DTC (Direct-to-Consumer) telehealth-plus-Rx (prescription). Nobody has built the audit-grade clinical platform underneath any of these. Click any player to see how they're positioned — and where Omnis sits relative to each.
Cluster · 01

Consumer DTC telehealth — US benchmark

Cluster · 02

Singapore — saturated, closed

Six players. Premium tier maxed out. Building a 7th here is not a venture — it's a confirmation that we read the market wrong.

Cluster · 03

India — consumer + metabolic

Fragmented. Each player owns one slice — wearable, generic pharmacy, longevity clinic, coaching. None has built the consolidated stack.

Cluster · 04

India — longevity (still embryonic)

Total Indian capital deployed in longevity 2016–2025: ~$50M across 12 rounds (Unisync Angels). Real category formation but pre-scale.

Cluster · 05

India — peptide / aesthetic clinics (regulatory gray zone)

Worth knowing the cluster exists, but not where Omnis plays. These operators sit in the legal-but-unenforced gap — Mumbai chains charge ₹20–40K/month for peptide stacks sourced through B2B "research peptide" wholesalers. The 24 March 2026 DCGI advisory (49-entity audit) signals the window closing. Omnis sells clean audit-grade substrate to whichever survivors operate post-enforcement; we don't build the consumer-facing peptide clinic ourselves.

The punch line

Where Omnis sits — the white-space.

Every player above owns one slice of the stack:

  • Consumer brand + DTC fulfilment · H&H, Ro
  • Annual biomarker labs · Function, Protocol Health, MORROW
  • Coaching app + wearable · Healthify, Ultrahuman, Twin Health
  • Pharmacy distribution · Truemeds, 1mg, Apollo 24/7, PharmEasy
  • Single-location clinic concierge · Biopeak, Sukino, the SG-saturated set
  • Aesthetic-front-end gray-channel peptide · ARQ, TAC, Mumbai chains

No one is the audit-grade clinical decision + monitoring + dispensing platform underneath all of them.

Omnis is that layer. The platform with on-device clinical AI, the gate pipeline, the watcher trio, and the provenance log — sold B2B to the clinics, telehealth platforms, and hospital networks that need to be HIPAA / PDPA / DPDP compliant by construction and to compete in the next 24 months as regulators tighten on the rest of the market.

The geographic question — which market we deploy first — is the optionality. The platform is geography-agnostic.

12

Where Omnis plugs into the institutional healthcare PE playbook

Four institutional-PE healthcare positions, each from a different angle. Omnis sits as the platform layer underneath: prescribing decisions, monitoring loops, audit-grade provenance. Click any position to see the integration mechanic in detail — what Omnis ships, what data flows back, what the partnership shape looks like.
The proposal

The shape we're proposing.

Institutional healthcare PE underwrites companies that build durable specialty-pharma + healthcare-services infrastructure across regulated markets. Omnis is the platform layer that sits underneath every clinical decision those companies eventually need.

A regional diagnostics distributor's lab volumes need an AI overlay. A specialty-pharma platform's products need a prescribing + RWE layer. A complex-generics platform's output needs post-Rx surveillance. The next hospital-services chain needs operational substrate that scales clinical quality faster than human bandwidth allows.

Omnis is one stack that does all of those.

The open question is which of the four positions (or an adjacent one not named here) is the right first integration — and what the partnership shape would look like.

13

What we want to find

Not a capital ask.Two specific things we're looking to settle next — the rest is direction and refinement.
01
Ask

Where Omnis plugs into the institutional healthcare PE playbook — customer, partner, or downstream operator?

A specialty-pharma platform is the freshly-closed shape. A regional diagnostics distributor is the next-to-cycle shape. A complex-generics platform is the mid-cycle shape. A hospital-services roll-up is the rear-view shape but the pattern repeats. Of those four — or an adjacent position — where the first integration actually lives matters: customer (Omnis platform licensed into their clinic networks), partner (co-built next operating instance), or portfolio operator (Omnis Digital Clinic runs a flagship inside the broader healthcare-services thesis, with the platform earning underneath every clinic on the network).

Which one comes first changes what gets shipped in the next 6 months and where Omnis Digital Clinic operates first.

02
Ask

An operating conversation with a regional diagnostics distributor — to test the AI-overlay-on-lab-volumes mechanic

The institutional buyer profile already holds positions in SEA diagnostics distribution — devices, lab equipment, IVD across the SG / MY / India / Indonesia / Philippines / Thailand corridor. Every lab result delivered through that footprint is a data point Omnis could interpret + protocol-recommend on. That's the most natural pilot partnership for the platform layer specifically.

A 60-minute conversation with the operating team of a comparable distributor would tell us whether the integration mechanic mapped here matches how distribution actually delivers value to clinic customers — or whether the real shape is something else.

We came in thinking consumer longevity. After watching specialty-pharma and healthcare-services capital close into the regulated-markets corridor, we think the right shape is something else. We're here to find it.

What we're certain on: the four-piece platform — gate pipeline, watcher trio, on-device clinical AI, frontier-compound readiness. The platform exists, it survives scrutiny, and it slots into the regulated-markets infrastructure thesis institutional healthcare capital is already underwriting.

What we're open on: geography, operating-co posture, business-model emphasis, partnership structure. Those answers come out of the right partnership conversations.

Schema
PatientRecord v1
EHR
FHIR R4 ready
API
/api/v1/patient
AI
Gemma 4 31B · on-device
Audit
Append-only log
OMNIS