Features in depth
The mechanisms behind each answer, described as they run in the code.
Pharma calculation library
A tested Python module loads into every run: ANOVA with Tukey HSD and compact letter display, Dunnett with GraphPad star notation, 4PL IC50/EC50 with 95% CI and extrapolation warnings, release kinetics ranked by AIC, f2/f1 with FDA/EMA applicability checks, ICH Q1E shelf life with poolability at α = 0.25, and Arrhenius t90. On its validation data, f2 reproduces the textbook value of 64.6.
In-place PDF editing
Text is replaced at the same position, size and colour, reusing the document’s own font when it has every glyph. Redaction removes the text instead of covering it. Forms, signatures, stamps, page numbers, watermarks and page operations are supported. After each replacement, the same search a PDF viewer uses must find the new text and not the old.
Workplace folders
Each folder carries a brief that a background stage rewrites after every chat: clients, prices, invoice numbers, deadlines, decisions and open items, each dated. Search is scoped to that folder alone, and folder rules hold: an email draft is never described as sent, and billing never invents a GSTIN or bank detail.
Deep scan, no model involved
A deterministic pass over the whole document checks percentages against their counts, arithmetic such as A ± B = C, mean ± SD recomputed from listed replicates, impossible p-values, and the same IC50, PDI, zeta potential or particle size reported with different values. Up to 200 findings, each marked ERROR or CHECK.
Checking and context
Deterministic checks, and the context that keeps a conversation going.
OCR and figures
Ten local OCR engines read scanned pages, including Hindi and Hindi with English, and every attachment is checked by its file signature, not its extension. Figures render in GraphPad Prism style with Arial-metric fonts, and only the final version of a re-rendered figure is shown.
Library and house skills
ACME and personal documents, including PDF, Word, Excel, CSV and images, are de-duplicated by content hash and indexed in overlapping chunks; spreadsheets keep sheet, row and cell references. Templates attach as working copies. Uploaded SKILL.md files, such as figure conventions or an SOP format, are followed by every channel.
What KRIYA reads, indexes and draws, and the rules it follows.
Knowledge and files
Request early access
Join the list for the hosted version of KRIYA Cortex, with sign-in for phone and tablet.
Contact
KRIYA Cortex is built by Dr. Akhilesh Vats at ACME Research Solutions, a GMP and ISO 9001:2015 certified pharmaceutical CRO in Baghpat, Uttar Pradesh.
info@kriyacortex.com
support@kriyacortex.com
© 2026 ACME Research Solutions. All rights reserved. Artwork generated in code from Indian structural motifs.