The data layer between a research plan and a successful experiment.
A reagent catalog tells you what exists. ABYLAY tells you what actually worked — which product a lab used for this exact experiment, what the paper really said in its methods, what happened when the lot arrived, and what it costs against the grant paying for it.
Open to verified university addresses. A six-digit code, no password.
Eight stages that normally live in eight different places.
A grant in a spreadsheet, an experiment in a notebook, a product on a vendor page, an outcome in someone's memory. ABYLAY keeps them on one thread, so a result can be traced back to the lot, the purchase, the price and the plan that produced it.
One product page. Three kinds of truth, never blended.
Manufacturer claims say what should work. Papers show where a product was used. Community outcomes say what actually happened on the bench. Averaging them into a single score would destroy the only thing that makes them useful — knowing which is which.
Published evidence
Peer-reviewed methods with the application, species and tissue the product was used in, and a link to the source every time.
A citation is never a verdictManufacturer data
Vendor specifications and validation, labelled as vendor-provided wherever they appear.
Never mixed with independent evidenceCommunity outcomes
Worked, partial or failed — with the exact conditions, the lot, and the failures that never reach a journal.
The layer that does not exist anywhere elseA missing number is a fact. A plausible one is a lie you cannot detect.
These are the product, not preferences. Every one of them costs a feature that would have looked better in a screenshot.
No invented figures
No fabricated prices, reviews, researcher counts or success rates. When a value is unknown the interface says so and explains why.
Unknown renders as unavailableMoney is arithmetic
Every financial figure is deterministic code over values a researcher entered or a verified vendor offer. No model touches a total.
One implementation, shared by browser and serverAI drafts, never decides
Extraction from a paper is a draft a researcher confirms or edits. Nothing a model produced is stored as a fact on its own.
The researcher is the author of the recordPrivate stays private
Lab history, grants and budgets are visible to members only. The database enforces it with row-level security, not the interface.
Policy, not a hidden buttonThe workspace opens with your university address.
Experiment matching, the product catalog, your lab's private history, grants and budgets — all behind a verified sign-in, because most of what is in there belongs to a lab rather than to the public.
The data layer between a research plan and a successful experiment.
Describe the experiment you are about to run. ABYLAY finds the products other labs actually used for it, the papers that back them, and what happened when the lot arrived.
Matching is deterministic: it narrows the catalog by the fingerprint fields above. It never invents a product, a citation or an outcome.
Four kinds of evidence, never mixed up
Also on every product page: the failure map, lot-to-lot comparison once the sample size supports it, and the grant budget line the purchase came from.
Today every step of this lives somewhere else: the grant in a PDF, the budget in a spreadsheet, the product choice in a Slack thread, the outcome in someone’s notebook. ABYLAY keeps them on one record, so the next person choosing a reagent sees what the last one learned.
One product page. Three kinds of truth.
Manufacturer claims tell you what should work. Papers show where it was used. Community outcomes tell you what actually happened in the lab — including failures that never get published. ABYLAY keeps the three separate.
Published evidence
Peer-reviewed methods, protocols, application, species and tissue context.
Provenance retained · a citation is never a “Worked”Manufacturer data
Vendor specifications and validation, explicitly labeled as vendor-provided.
Never mixed with independent evidenceCommunity outcomes
Worked / partial / failed results, exact conditions, lot notes and researcher comments.
The proprietary layerStart with your .edu address
We send a one-time code. No password to create, and none of your institutional credentials ever touch ABYLAY. Currently open to UW researchers.
Abylay Karagayev
Building the data layer between a research plan and a successful experiment — combining published evidence, real experimental outcomes, and the context that actually matters at the bench.
Evidence, not marketing copy.
Loading catalog…