Infectious disease co-pilot · pilot program

The right antibiotic,
the right dose,
every time.

Demed guides any doctor — not just infectious disease specialists — through medication selection, dosing, and de-escalation for infection cases. Every recommendation is grounded to the guidelines your hospital already follows, cited line by line. Not a chatbot. A clinical instrument.

app.demed.in/queue
NEEDS ACTION 4
CH-10241
68F · ICU-2
PENICILLINEMPIRIC
CH-10260
72M · ICU-1
SULFAADMIT
CH-10255
54M · MW-1
CULTURE
10.4L
Deaths associated with AMR in India — 2019 data, 2024 GRAM estimate
Lancet GRAM Project, Sep 2024
58,000
Newborns who die from antibiotic-resistant sepsis in India, every year
ICMR AMR Surveillance Network
56%+
Of India's AMR burden originates inside healthcare settings, not community spread
ICMR AMRSN Annual Report 2024
The problem

AMR isn't the disease. It's the symptom of a harder problem doctors face on every ward round.

Any physician — cardiologist, surgeon, internist, not just an infectious disease specialist — can end up choosing an empiric antibiotic under time pressure, with an incomplete picture of local resistance patterns, renal function, allergy cross-reactivity, and which of a dozen overlapping guideline documents actually applies to this patient, right now.

Get that choice wrong and two things happen: the patient's treatment is delayed or suboptimal, and every inappropriate or overly broad antibiotic prescribed adds to the resistance burden the next patient will face. AMR is the population-level scar left by thousands of individually rushed decisions.

Not another chatbot

Demed doesn't generate answers. It retrieves and cites them.

Ask a general-purpose LLM a clinical question and it will answer — fluently, confidently, and sometimes wrong, with no way to check its work. Demed is architecturally different: a deterministic rules engine handles anything calculable (renal dosing, allergy cross-reactivity) and never guesses, while every guideline-based recommendation is retrieved from a curated corpus of IDSA, ICMR, and WHO source documents and shown with its citation, not asserted from memory.

01

Deterministic where it must be

Dose calculations and allergy blocks run in code, not a language model — never a guess where a wrong answer has real consequences.

02

Every claim is cited

Recommendations show which guideline, which year, which section — not "trust me."

03

India-specific, not translated

Grounded in ICMR's own national resistance surveillance data — not just Western guideline assumptions.

How it works

From admission to de-escalation, in the same workflow you already run.

app.demed.in/empiric
Recommendations require physician confirmation and clinical-lead validation.
AdmitEmpiricCultureDoseMonitor
EMPIRIC THERAPY · RECOMMENDED
Meropenem
1 g · IV · q8h · 7 days
WATCH
CONFIDENCE
Guideline
85
Local
70
Patient fit
80
Cited: ICMR 2024 · IDSA 2010 · India AMR surveillance

Empiric therapy

One recommendation, one primary drug, full confidence breakdown — never a list to sift through under pressure.

app.demed.in/culture
DE-ESCALATION OPTIONS · NARROWEST FIRST
Nitrofurantoin
per protocol
ACCESS
Cotrimoxazole
per protocol
ACCESS
Narrowed from Cefepime (Watch) · E. coli susceptible

Culture-directed de-escalation

Once cultures return, Demed re-ranks toward the narrowest effective agent — Access before Watch before Reserve.

app.demed.in/ask
ASK DEMED
What is the empiric therapy for complicated UTI?

According to the IDSA 2025 complicated UTI guideline, empiric therapy depends on a four-step process: (1) assess illness severity, (2) evaluate risk factors for resistant organisms, (3) assess patient-specific factors, (4) consider the local antibiogram. Nitrofurantoin and oral fosfomycin are explicitly excluded — they don't achieve adequate levels in renal parenchyma or blood.

IDSA 2025ICMR 2019

Ask Demed

Any clinical question, answered from the guideline corpus with citations — not a free-floating LLM opinion.

On the roadmap

Voice input and multi-page document upload (lab reports, discharge summaries) are in active development for the next pilot phase — we're working on it and will update pilot participants directly as each capability goes live.

Before you sign up

Questions every doctor asks first

Does Demed replace my clinical judgment?

No. Every recommendation requires physician confirmation, and pilot-phase recommendations additionally require clinical-lead sign-off before they're treated as validated. Demed is a co-pilot, not an autopilot.

Where does patient data go?

Data is processed with tenant isolation per hospital and an immutable audit trail on every action. We're building toward full DPDP (India's Digital Personal Data Protection Act) compliance as a first-class requirement, not an afterthought.

What if Demed is wrong?

Every recommendation shows its confidence and its sources so you can judge it, not just accept it — and the deterministic safety logic (allergy, renal dosing) is designed to fail closed: no safe answer means no recommendation, never a guess.

Read the full FAQ →

Bring Demed to your hospital.

We're running a focused pilot with hospitals ready to validate a guideline-grounded co-pilot on real cases.

Request a pilot