Treat with the whole history in front of you.
PulseNote is an AI powered EMR for doctors in Bangladesh. OCR brings in the paper records you already have, and Pulse AI reads the record with you.
We started where a mistake is not recoverable.
Children with cancer. Growth curves that change the treatment, drugs that are dangerous at the wrong dose, and a history nobody can afford to guess at. Twice a day the ward changes hands, and the doctor coming on used to read back through the whole file to work out what the last one did and why. That is now one handover screen. We did not start there because that is the product. We started there because an EMR that holds up in that department will hold up in your chamber. This is a pilot, not a signed deployment, and we will not describe it as anything else.
The visit writes the record. Not the other way round.
A patient sits down and you open them. The last visit is already on screen, with the diagnosis, the prescription and what you advised. You fill in the vitals, the complaint and the history, choose the diagnosis by searching in English or Bangla, pick the drug by the brand name sold here, and print the prescription. Writing the prescription is what wrote the record, so nothing is entered twice.
Diabetes mellitus
Comet · 1+0+1 · 30 days
Return in 8 weeks with fasting blood sugar.
Search by disease name or বাংলা নাম
A fresh visit record has been saved and linked to their history.
Download PDFScripted sample data. The console replays when it scrolls into view.
Every EMR is empty on the day you buy it.
That is why they fail. A doctor switches, the years of history do not come with them, and the thing is useless for months until enough visits pile up. Nobody is going to type the backlog in, so we do not ask. Photograph what the patient already carries, the old prescriptions, the lab reports, the discharge summary from the hospital across town. That is what OCR is for, and it is why the record is worth reading in the first week rather than the sixth month.
Pulse AI finished reading the document.
Read as a prescription. What opens next follows what it actually found on the page.
Opens the review screen with every medication it read, ready to check.
The shared screen understands the clinical-note shape, which is what this was read into.
Napa, Omeprazole, Zithrox. Added from a photograph, confirmed by Dr. Adnan.
Vincristine and dexamethasone given. Counts reviewed.
ANC 0.4, Hb 7.2, platelets 38.
Tap a document type to run another one.
Still on paper? Digitised in minutes.
Point a phone at any medical document. Prescriptions, lab reports, imaging, discharge summaries, referral letters, clinical notes, history sheets, round notes. Pulse AI reads it, routes it to the right form, and a doctor confirms it before anything enters the record.
The whole chart, read before you reach the bed.
Pulse AI is not a chatbot bolted onto a records app. The record was built for it to read. Every visit, every result, every page you photographed. Ask in the language you already speak and it answers from that child’s chart, and shows you where the answer came from.
Counts have not turned yet.
- 6 Apr1.1L
- 8 Apr0.9L
- 10 Apr0.7L
- 12 Apr0.4C
- 2 Apr9.8L
- 6 Apr8.9L
- 9 Apr8.1unreviewed
- 12 Apr7.2L
- ·The neutrophil count has fallen at every reading this week.
- ·Haemoglobin is below the range for this age and still falling.
- ·A photographed CBC report was read and confirmed against the chart.
- ·No blood culture result is on file yet.
- ·The next set of counts is due tonight.
- ·The diagnosis and the treatment protocol are unchanged.
- No blood culture result is on file.
Weight has been flat over the same period, which may be worth reading alongside the counts rather than on its own.
Anchored in: Weight, ANCThe list is unchanged since the last round.
Nothing has been started or stopped since the last review.
- Vincristine 1.5 mg/m²IV, weekly, started 12 AprActive
- Dexamethasone 6 mg/m²oral, BD, started 2 AprActive
- Co-trimoxazole 240 mgoral, BD, started 2 AprActive
Pulse AI can make mistakes. Verify clinically critical answers.
Invented sample data. Tap a command to ask.
The dangerous moment is the dose.
Four things sit one tap from the record, so nobody has to leave it to check.
- Drug reference
- Built on Bangladesh's own DGDA registry, so it holds the brands actually sold in this country and the names patients actually bring in, not an imported database of products nobody here can buy.
- Diagnosis search
- Type in English or Bangla and pick the diagnosis. Both spellings of the same condition find it, which is what makes a searchable record possible in a bilingual clinic.
- Dose calculator
- Per kilogram, against the weight already in the record. Built for children, where the wrong dose is the dangerous one.
- Your own protocols
- Upload your unit’s guidelines and Pulse AI answers from them, naming the passage it used. Answering from a document your department already agreed on is different from answering from the internet.
The hardest department we could get into.
We are piloting in paediatric haematology and oncology at Bangladesh Medical University. Children with cancer, thirty visits deep, growth curves that carry clinical meaning, and drugs that are dangerous at the wrong dose. If a record holds there, it holds anywhere. Everything below is built and in the pilot.
Round to round
- Shift handover, end to end
- The roster for who is on, an inbox for the incoming shift, a record of everything passed on, and a printable sheet to carry on the round.
- Ward census
- Every bed by stage, ward or bed number, with who is critical readable at a glance.
- Department dashboard
- Census, throughput, length of stay and outcomes, for the unit head.
- Roles that match the unit
- Resident, supervisor, department head and admin, each seeing their own view.
The record, end to end
- Admission to discharge
- One timeline per child, including the hospital course and the discharge summary.
- The unit’s own forms
- Visit notes and daily round notes on the paper forms the ward already uses.
- Growth against WHO standards
- Weight and height for age scored and charted across every visit.
- Labs that trend
- Counts followed across admissions, with the direction called out, not just the latest value.
Paper, turned into data
- Photograph anything
- Prescriptions, labs, imaging, discharge summaries, referrals, history sheets, round notes.
- A review queue
- Everything read from paper waits for a doctor to confirm it before it counts.
- Filled forms, not free text
- What was read lands in the right field of the right form, ready to check.
The record is theirs. That is why it reaches you at all.
Every record you write belongs to the patient as well as to you. It lands on their phone in an account they own rather than one a hospital owns. That is not only a principle, it is the mechanism: because the record travels with the patient rather than sitting in one clinic’s registry, the history from a hospital you have never dealt with is on your screen when they walk in.
- It comes back to you
- The patient who returns after a year arrives with the history already there. Not a bag of paper, not a guess, not a test repeated because nobody could find the first result.
- The family in one account
- A mother manages her child’s record alongside her own and her parents’, each one kept separate, each one hers to share or withhold.
- They can add to it
- A patient can photograph their own lab report between visits. It runs through the same OCR and waits for someone to check it, exactly like anything you upload.
What the department head sees on Monday.
The same EMR that makes a consult better makes a department legible. A ward census by bed and stage, shift handover from the outgoing team to the incoming one, admission through to discharge on one timeline, and four roles that each see a different view. There is no separate reporting exercise, because it is the same record the round already wrote.
- A census that is current
- Every bed by stage, with who is critical readable at a glance. Not a count somebody assembled on Monday morning from a register.
- Handover that leaves a record
- Who is on, what was passed to them, and a printable sheet to carry on the round. Nothing important survives only in the outgoing resident’s memory.
- Admission to discharge
- One timeline per patient, from registration through the hospital course to the discharge summary, with throughput and length of stay falling out of it rather than being compiled.
- Four roles, four views
- Resident, supervisor, department head and admin. Your department decides who is who. A second department gets its own configuration, not a second copy of the system.
- Specialty measures, where they matter
- A paediatric unit gets weight and height scored against WHO standards on every measurement, and an alert when a child crosses a critical threshold. A department that does not need that does not see it.
The law changes in 2026. This was built for it.
Bangladesh's Personal Data Protection Act commences in 2026. Nobody is certified against it yet, ourselves included. So here is what is actually in the system today, and what we will not claim.
PDPA 2026
Bangladesh's Personal Data Protection Act. We built to it before it commenced rather than retrofitting after.
Encrypted
Clinical data is encrypted in transit and at rest. We do not claim end to end encryption, because we do not have it.
Audit Trail
Every access to a patient record is written to a log the application cannot edit or delete. Not by a doctor, not by an admin, not by us.
Role-Based Access
Resident, supervisor, department head and admin each see a different view. Your department sets who is who, not us.
FHIR Compatible
Built on the global standard for exchanging health records, so a record can move to another system. Compatible, not certified.
We built every layer, because a national health record has no shortcut.
The record, the OCR that fills it, the AI that reads it, and the standards that let it leave. We built all four rather than assembling them, because in Bangladesh there was nothing underneath to assemble. That is slower, and it is the only way the infrastructure ends up belonging to the country rather than to a vendor.
Built and running today
- The record
- Visits, prescriptions, lab reports, discharge summaries, one timeline per patient, owned by the patient.
- OCR
- Any medical document read straight into the record, with a person checking it before it counts.
- Pulse AI
- Reasoning over the record, checked before it reaches the doctor, with a hard boundary on what it will say.
- Standards
- FHIR R4 export and HL7 v2 import, so a record can move to another system.
On the roadmap, not yet built
- Deeper clinical reasoning
- Pulse AI will reason further against uploaded national and departmental guidelines, not only the record.
- More hospitals
- A second and third department will be set up without new code being written for them.
- Population health
- De-identified analytics for disease surveillance, so that an outbreak would become visible while it is still small.
- National reporting
- Export to DGHS, and a blood donor registry that could fan a request out across districts.
The endgame is a country where a health record is portable, and where a doctor anywhere in it starts from something rather than nothing.
Start with one clinic.
Not a rollout and not a procurement cycle. Your own patients, your own drugs, your own way of writing a prescription. If you run a department, we set it up on your forms, load your protocols into Pulse AI, and sit in your rounds until it works.
We started in Bangladesh because that's where we're from and that's where we know. Tell us where you practise.