The clinical-AI appliance · SBX Group, Tokyo

Time to think.Time to care.

Disha is a sealed clinical-AI appliance that sees, hears and speaks. Its agents read what the ward produces — prescriptions, case sheets, bedside monitors, HIS records — draft final documents, surface the clinical picture, and hand everything to your clinicians to validate. No cloud. No new logins. No data ever leaves the room.

Happy providers make happy patients.

Request a ward demo See the clinical evidence
Clinician-firstAccreditation-readyPlug & play
The Disha Box prototype, photographed July 2026 — a white swan-neck clinical appliance with its dedicated AI engine in the base
0 logins added
≤12B params, on-device
0 bytes to the cloud
The Disha Box · Prototype · July 2026
~3.5 hrs
clinician documentation effort returned per day, at ward scale
Chennai pilot · endpoint report
2–10+ min
saved per patient review by the Report of Reports agent
Chennai pilot · RoR trial
~60% faster
discharge summaries in a live floor test at a Kolkata hospital
Dishari field immersion · 2026
30+
hospitals engaged across South-East Asia & APAC
Dishari lighthouse program
Why Disha · bridging the EHR gap

The systems are digital. The work is not.

Care is DUST(y) — and two decades of digitisation added logins, forms and click-paths without taking the tedium away. Disha automates the data entry that bottlenecks clinical operations, so the EHR you already own finally gets adopted — and your clinicians get room to be human in the difficult, uncertain, scary parts.

DifficultUncertainScaryTedious
  • OPD10+ page paper records reconstructed by hand at every consultation
  • WARDNurse handovers carried verbally and in handwritten registers between every shift
  • ADMITRepeat patients re-registered from scratch — history never auto-fills
  • ICCUDisproportionate time on handwritten notes and redundant verification
Field observations · Dishari program, India 2025–26

What the manual workflow costs

Baselines measured in pilot hospitals, before Disha
One discharge summarywrite · type · format, across shifts
6–8 hrs
One discharge summary — Kolkata floor testobserved live, side-by-side
15–20 min
One handwritten prescriptionread · decode · type into records
2–3 min
Measured baselines · Dishari program · partner hospitals, Chennai & Kolkata
Disha Box prototype, photographed on the bench — the dedicated AI engine and local-only I/O panel seated in the champagne housing
Prototype unit, as photographed · dedicated AI engine · fanless & silent
The product

The Disha Box.

A sealed clinical-intelligence appliance with minimal infrastructure demands — plug & play. Agents arrive preinstalled — no additional infrastructure required — and it sits beside your HIS/EMR, not in place of it.

零

No data ever leaves the room

Every model runs on the box. Raw inputs never leave the device — data residency is a property of the hardware, and ownership rests with your institution.

無

No cloud. No internet required

Fully offline operation at every stage — privacy-preserving and designed for compliance by construction.

直

No new logins or dashboards

Agents learn your workflows; your staff never learn new software.

2048-coreAI GPU
64 GBLPDDR5 RAM
4 TBNVMe SSD
≤12Bparams on-device
How it works

How it works — from capture to a validated record.

Inputs: paper documents, bedside monitors, room audio and HIS records. Outputs: consultant-ready draft documents and structured clinical reads — every output validated and signed by your clinicians. The HIS is a source as well as a destination: RoR reads investigations back into trends and a provisional read.

Diagram of the Disha loop inside a dashed boundary labelled ‘your ward’: prescriptions, case sheets and bedside monitors are photographed in place; the Disha Box — which sees, hears and speaks — produces drafts and clinical reads entirely on-device, ending in final consultant-ready documents; the clinician validates and signs as final authority; the signed record lands in the hospital’s HIS or EMR — and a dashed return arrow shows RoR reading investigations back from the HIS into trends and a provisional read. Zero bytes leave the room — no cloud, no new logins.
Autonomy is a slider, not a switch — human validation remains the final authority on every output.
The Disha Agency

Six task-specific agents — in production today.

Interoperable, domain-limited, and immersed in your existing workflows — from Prescription Readers to Discharge Summary Generators, deploy the intelligence your hospital needs, on-site.

Agent · 01

Prescription Reader

Handwritten and printed scripts into structured, verifiable medication records — low-confidence fields flagged for human review, tuned to local formularies and abbreviations.

Agent · 02

Clinical Notes Reader

Case sheets, progress notes and mixed-format documents into structured summaries the care team can act on.

Agent · 03

Measurement Monitor Reader

Vitals straight from photographs of bedside monitors and instruments — one nursing round, complete structured chart, no re-typing.

Agent · 04

Report Generator

Clinical and operational reports assembled from agent outputs, with transparent sourcing back to the underlying documents.

Agent · 05 Flagship · live in Chennai

Discharge Summary Generator

Consultant-ready discharge summaries from the full case sheet — clinically deployed at a 200-bed multi-specialty hospital in Chennai, with published endpoints.

Agent · 06 Trialed in Chennai

Report of Reports (RoR)

Investigations from HIMS/LIS into trends, abnormal summaries, hospital-format charts and a grounded provisional diagnosis.

+ Radiology Explainer · Trial Organizer · Data Harmonizer · De-identifier · voice productivity agents — the full catalogue lives in the agent showcase.

Clarity & consistency

Domain-limited, real-world problems with low-friction, measurable scope.

Transparent reasoning

Background by default, surfacing when needed. The clinician always sees why an output says what it says.

Controlled autonomy

Every agent declares intent and reveals action. Human validation remains the final authority.

Early traction · measured before we scale

Deployed. Measured. Published.

Two hospitals in, measured before we scale — a 200-bed multi-specialty hospital in Chennai and a working hospital in Kolkata. Documentation was the first measured endpoint; the same on-box loop powers all six agents. Disha is deliberately evidence-driven: deploy against the manual workflow, measure clinical endpoints, publish the result, and let you inspect it before you buy.

“AI must be one integrated experience — not disconnected plugins. Human validation is mandatory.”
Senior Consultant, partner hospital, Kolkata — field immersion, May 2026
Clinical endpoint report · Chennai pilot

Dishacharge — the discharge-summary agent, live on inpatient wards

~3.5 hrs/day
clinician documentation effort saved across 27 planned discharges — a ~50% reduction per case, at 85%+ generation quality

Endpoint-style measurement of capture, processing, validation and generation quality at a 200-bed multi-specialty hospital — ~14 minutes end-to-end on a 25-page case sheet.

Full endpoint report — available on request
IPD evaluation · Chennai pilot

Report of Reports, judged by duty medical officers

2–10+ min
saved per patient, per review — six features assessed in the initial three-DMO evaluation

Investigations pulled from HIMS/LIS and turned into trends, abnormal summaries, hospital-format charts and a grounded provisional diagnosis.

Full trial report — available on request
Field immersion · Kolkata

Six days inside a working hospital

~60% faster
discharge summaries in a live floor test — 15–20 minutes down to ~7, reviewed and signed by the clinician

Thirteen stakeholders, two facilities, 201 artifacts catalogued — closing with an HDU-floor pilot authorized by hospital leadership.

Read the field immersion report →
Metrics are from ongoing pilots under controlled trials; performance varies with workflow design, documentation quality and clinical complexity.
More on the way

A pipeline shaped by the field: Prescription Reader hardening, voice capture for OP consults, handwriting OCR, insurance workflow assistance — and engagements in discussion, e.g. a behavioral-health edition of the box with a regional wellbeing group, drafting session documentation in the language of the room. In discussion — not yet deployed.

The Yutori calculator

How much time would your wards get back?

Modelled on Disha’s published pilot measurements for discharge documentation. Move the slider to your hospital’s planned daily discharges. Discharge is simply the first endpoint we’ve published — one agent of six.

3.5 hrsclinician time returned per day
105 hrsreturned per month (30 days)
14additional 15-minute consultations possible per day
≈ 2.6full 8-hour clinician-shifts returned per week
Basis: ~7.8 min saved per discharge (Chennai endpoint report — ~3.5 hrs across 27 discharges/day).
Clinician-first · Accreditation-ready

Accreditation-ready, by design.

Disha is being built for the evolving regulatory landscapes of South-East Asia — NABH (India), GDAC (Cambodia), and the frameworks of Singapore and Malaysia — with the documentation completeness, consistency and audit trails that earning and keeping your hospital’s accreditation demands.

It is engineered to comply with regional health-data protection standards: every model runs on the box, so data residency is a property of the hardware, not a promise in a contract. And because every output is reviewed and signed by your clinicians, Disha works today as a clinician-assist layer — human validation is the final authority.

Designed for the requirements of
NABHIndia JCIInternational GDACCambodia MOHSingapore MOHMalaysia
Formal regulatory certifications are progressing as part of our roadmap — we don’t claim them before they’re earned.

Private by construction

Not policies — properties of the hardware.
  • 零
    Every model runs on the boxUp to 12B parameters on-device — raw inputs never leave the appliance.
  • 無
    Fully offline, at every stageNo cloud, no internet required — data residency is physical, not contractual.
  • 直
    No new logins or dashboardsAgents join your workflows; nothing new to secure, train, or audit.
  • ✓
    Human validation is finalEvery output is reviewed and signed by your clinicians before it counts.
How you deploy

Start with one ward. Scale on evidence.

Three deployment models, one sales ladder — and every step measured before you take the next one.

Model · Bed

Bed

Per-bed on the ward — the discharge desk, the nursing station, the ICCU corner. The smallest footprint with the fastest payback.

Model · Bench

Bench

A per-department workbench — radiology, pathology, the OPD — with the agents that department actually needs.

Model · Combini

Combini

The whole-clinic bundle for smaller facilities — one box, all six agents, everything on-site.

The ladder Veni — pilot ward  →  Vidi — department  →  Vici — enterprise.  Pricing & ROI simulator available for CFO and procurement teams.
De-risked by Dishari

Every deployment can begin as a measured Dishari pilot — the agent runs against your manual workflow, clinical endpoints are recorded, and the findings report is yours to keep and inspect before you scale. It’s how every number on this page was earned.

Live and commencing across Kolkata · Chennai · Bengaluru · Kuala Lumpur · Singapore · Tokyo · Osaka
Why trust us

25 years of systems science — made in Japan, for the world.

Disha is the clinical-intelligence product of the SBX Group — the digital-biology company spun off in 2011 from the Systems Biology Institute, Tokyo, the non-profit that helped create the field of systems biology itself. Bootstrapped, profitable and stable for over a decade, and powered by the Garuda agent framework.

25 yrsof innovation from SBI Tokyo
35+partners across pharma, healthcare & medtech
Top 3pharma companies in Japan served
200+peer-reviewed research publications
F500global consumer-health clients
17+countries in the research network
Garuda connects · Taxila mines knowledge · Gandhara learns — Disha is where the platform stack meets the clinical encounter.
Leadership
Portrait of Dr. Hiroaki Kitano
Dr. Hiroaki Kitano

CEO, SBX Corporation · President, Systems Biology Institute Tokyo · CTO Fellow, Sony Group

Portrait of Dr. Samik Ghosh
Dr. Samik Ghosh

COO, SBX Corporation · CEO, SBX Technologies · PhD CS, University of Texas

Portrait of Dr. Sucheendra Kumar Palaniappan
Dr. Sucheendra K. Palaniappan

VP, SBX Corporation · CTO, SBX Technologies · PhD CS, National University of Singapore

Bring Disha to your ward

See it working — on your own floor.

A ward demo takes one desk, one afternoon, and none of your patient data. If it earns its place, start a measured pilot — the endpoints are recorded, and the findings report is yours either way.

Request a ward demo Pricing & ROI simulator
Investors & strategic partners — request the data room · ghosh@sbx-corp.com · suchee@sbx-corp.com · +81 (0)3-6456-4250