§ 01 — RESEARCHBERRY OVERVIEW
Mayavis · ResearchBerry — No. 01

Build research, not formatting.

Your personal AI research editor. Paste the rough notes or the thesis idea. ResearchBerry asks the pointed questions, finds and verifies the citations, drafts the structured paper, and takes every edit in plain English — then hands you Word, PDF, or LaTeX.

from: rough_notes.txtto: manuscript_v1.tex■ MULTI-VENUE READY
ACADEMIC RESEARCH JOURNAL · 2026 EDITION

LLM-Assisted Triage in Emergency Departments: A Convergent Mixed-Methods Evaluation

M. Viswanathan1, ResearchBerry AI2
1Mayavis Research Studio  •  2Citation & Style Engine
Abstract—Evaluating LLM-assisted triage across three emergency departments (n = 312). ResearchBerry verified 23 citation DOIs, structured sections §1–§6, and formatted the manuscript for IEEE, APA, Chicago, LaTeX, Word, and PDF.
ObjectiveED Triage • n = 312
Citations23 Verified DOIs
StylesIEEE • APA • Chicago • ACM
ExportsLaTeX • PDF • Word
FORMAT READY ✓
No. 01 — First product of Mayavis Private Limited Scroll — the notes assemble themselves
( 01 ) — The product

Watch a paper think.

Not a screenshot. A faithful simulation of the ResearchBerry flow — paste, ask, verify, draft, export — compressed into a single scroll. Keep going; the manuscript writes itself.

ResearchBerry — live simulationACT I — PASTELive
rough_notes.txt — 11:47 PM

context parsing…
01 — Paste. Messy is exactly where it starts.
RB

What outcome defines successful triage in your setting?

YOU

Shorter time-to-first-assessment — without missed deterioration.

RB

Which method carries the primary claim?

YOU

Quantitative. The interviews explain the why.

RB

Three hospitals — comparable data?

YOU

Two on Epic, one paper-based. Site becomes a covariate.

3 questions · 41 seconds · context built
02 — Ask. It builds context before it builds sentences.
LLM triage Chen 2024Okafor 2023 Rao 2025Esteva 2021 WHO 2022
Chen et al. · JAMIA 2024
LLM-assisted triage in emergency care: a systematic review.
DOI 10.1093/jamia/ocae031 ✓ Verified
Triage guideline · 2016
Older protocol — superseded by 2022 update.
Dismissed
Rao & Iyer · npj Digit. Med. 2025
Mixed-methods evaluation of decision support in EDs.
DOI 10.1038/s41746-025-0142 ✓ Verified
03 — Verify. Every citation traceable. None invented.
Large Language Models for Emergency Department Triage: A Mixed-Methods Study
Draft v1 · 6 sections · 23 citations

We examine whether LLM-assisted triage reduces time-to-first-assessment across three emergency departments, while preserving detection of clinical deterioration.

We looked at some data and also talked to a few clinicians about it.
Clinical notesDe-identifyLLM triageRisk scoreClinician feedback loop
04 — Edit. The draft obeys your sentences.
Large Language Models for Emergency Department Triage: A Mixed-Methods Study
Submission draft
We examine whether LLM-assisted triage reduces time-to-first-assessment across three emergency departments while preserving detection of clinical deterioration. A convergent mixed-methods design pairs a retrospective cohort analysis with semi-structured clinician interviews. Model-assisted triage reduced median time-to-assessment by 22% without increasing missed-deterioration events. Clinicians described calibrated trust as the mechanism behind sustained adoption.
[1] C. Chen et al., “LLM-assisted triage,” JAMIA, 2024.
[2] A. Rao & S. Iyer, “Mixed-methods evaluation,” npj Digit. Med., 2025.
[3] WHO, “Emergency triage guidelines,” 2022.
manuscript.docx manuscript.pdf main.tex
05 — Export. Word · PDF · LaTeX. Your venue’s style.
( 02 ) — The method

Four moves. No ceremony.

02 / 06
01

Paste your ideas.

Drop the thesis, the scribbles, the two-in-the-morning paragraph. Messy is the point — ResearchBerry begins exactly where you actually are, not where a template pretends you are.

02

Answer smart questions.

Before a single sentence, it interviews you — the gap, the method, the claim you are willing to defend. Context, not assumptions. The draft that follows already knows your study.

03

Choose verified citations.

It searches the literature and shows its work — every paper real, every DOI checked against its source. You approve what gets cited. Nothing phantom slips into your bibliography.

Chen et al., JAMIA 2024 ✓ Verified
Rao & Iyer, npj 2025 ✓ Verified
WHO guidelines, 2022 ✓ Verified
04

Edit in plain English.

“Rewrite the introduction.” “Add a limitations section.” “Make it IEEE.” Said, done — and every dependent section follows the change without being asked twice.

The results were good.
Effect sizes were large and consistent across all three sites.
( 03 ) — Why it matters

The ritual ends. The thinking stays.

03 / 06
0
hours of mechanical work returned to the work that actually matters — per paper, on average during beta.

The ritual

  • 01Export citations to a reference manager
  • 02Re-export them when the manager corrupts the file
  • 03Chase down broken DOIs by hand
  • 04Hunt for PDFs behind paywalls
  • 05Reformat the bibliography — again
  • 06Convert the document from Word to LaTeX
  • 07Fix the LaTeX errors the conversion invented
  • 08Resize diagrams until they fit the column
  • 09Reformat everything for IEEE
  • 10Rewrite the abstract for the seventh time
  • 11Check the margins against the style guide
  • 12Rename the file final_final_v7

The work that matters

  • Thinking the question through.
  • Reading deeply, not skimming abstracts.
  • Deciding what the evidence actually says.
  • Discovering the thing no one asked you to find.
( 04 ) — Capabilities

Everything a manuscript needs. Nothing it doesn’t.

04 / 06
( 05 ) — The company

Building the future of AI-powered productivity.

Mayavis makes intelligent software that amplifies human capability. Not to replace researchers — to remove the repetitive work that keeps them from researching. The thinking, the discovering, the deciding: that stays human. Everything that orbits it becomes software, quietly, in the background.

  • 01Research · ResearchBerryAvailable now
  • 02Enterprise productivityIn development
  • 03Document intelligenceIn development
  • 04Knowledge workOn the horizon
ResearchBerry workspace — monitor displaying AI research manuscript with equations, verified citations, and research notebook on pale stone desk. Mayavis Research Studio — Noida, UP, India

“The company is bigger than the product. The product introduces the company.”

( 06 ) — Begin

Ready to experience the future of research?

Try ResearchBerry
Free during beta · No credit card · Your notes stay yours