SovereignMalayalam AI.
MALAYALAM AI[01]
OPEN SOURCE & STUDIO[02]
THE COMPANY[03]
· About Araskova Labs
The data, benchmarks and models Kerala runs on its own infrastructure.
A founder-owned, fully remote team in Kerala. We measure how leading AI models fail on real Malayalam tasks, build the licensed corpus and benchmarks to fix it, and deploy models that run on-premise.
കേരളത്തിൽ നിർമ്മിച്ചത്
Built in Kerala
· 002 — The Lab
We measure where leading AI fails on Malayalam, build the licensed data and benchmarks to fix it, and ship models that run on the institution's own hardware.
Measured
Every claim carries the date we measured it.
On-premise
Models run on hardware the institution owns.
Studio-funded
Client work pays for the research.
· 003 — The problem
General models learn mostly from English and run from someone else's cloud. Kerala's institutions need models that read Malayalam properly and stay on their own hardware.
· 004 — Method
Four stages, the same for every system the lab builds: collect, measure, build, deploy.
· 005 — From the Lab
Built on our Malayalam corpus and benchmark, and deployed on the institution's own servers.
· 006 — Research
What the lab is working on, with the figures we've measured and the date we measured them. Each track links to its research note.
01 / 04 — Stepping through
~73%
FAILURE RATE · 4 MODELS · JUNE 2026
200 real-world Malayalam administrative queries, scored against four leading models in June 2026. Failure means code-switching, transliteration errors, semantic drift or invented terminology.
· 007 — Who it's for
Malayalam AI is for everyone who reads and writes it, from a student to a district office. Some of it we publish openly; some we build and install for an institution.
AI that answers properly in Malayalam, not a translation of English. We test leading models on real Malayalam and publish what we find, so people know which tools they can trust in their own language.
Our research →Research notes on Malayalam evaluation, corpora and OCR, published as we measure, and a research fellowship for people who want to work on them with us.
Research fellowship →Open-source tools anyone can use and build on, released every so often under open licences.
Open source →Read Malayalam files, circulars, registers and case records, and answer staff questions from them, on the department's own servers.
Talk to us →Turn handwritten Malayalam forms and ledgers into structured data that stays inside the bank.
Talk to us →Digitise Malayalam patient records and referral letters on hardware the hospital controls.
Talk to us →Make printed and pazhaya lipi collections searchable, with the source and licence of every document recorded.
Talk to us →Custom software and AI automation, built by Araskova Studio with direct builder access and no agency markup.
Araskova Studio →· 008 — Accepting select commercial & research contracts
Your records, your hardware, and what the system has to do in Malayalam: that's all we need to start.