Company · Sokros Labs
Frontier work, in open files.
Sokros Labs is where the layer meets problems ahead of the product: research lines, national-scale scoping, public-interest casework. Each one is kept as an open file that says who it's with and how far it has got.
One file live today: CIPD marking across 34 centres in 20 markets
01 · Open files
The ledger, kept honestly.
Live deployments, research lines and programmes being scoped, each with its status on record. Where the work is early, the ledger says so.
PRG-01
Regulated qualification marking
CIPD units at Levels 3, 5 and 7, marked in production across 20 markets.
PRG-02
Deterministic evaluation research
With Google DeepMind: how far a judgement pipeline can be verified with no probabilistic step in the loop.
PRG-03
Education infrastructure for new cities
Assessment rails scoped for NEOM's education programmes: marking, verification and audit as one layer.
PRG-04
Arabic-first extraction
With MBZUAI: the extraction stack is multilingual today; this makes Arabic a first-class marking language, not a translation.
PRG-05
National skills frameworks
Alignment work with SDAIA and UAE education authorities: one deterministic record beneath national qualification programmes.
PRG-06
Government casework decisions
Scoped with UK public bodies: licensing and casework queues judged against published criteria, with the full rule trail on every decision.
02 · The partners
The company we keep.
Research groups, regulators and the bodies whose names are on the certificate each hold a row here. A name joins the wall when an agreement is signed, not before.
Research
evaluation methods & extraction science
- Google DeepMind
- MBZUAI
- The Alan Turing Institute
Government & regulators
education and AI oversight, UK · GCC
- Department for Education
- Ofqual
- UAE Ministry of Education
- KHDA
National programmes
AI & education initiatives, KSA · UAE
- NEOM
- SDAIA
- HUMAIN
- G42
Professional bodies
qualifications marked on the record
- CIPD
- CMI
- ACCA
- Pearson
Centres, awarding bodies and platforms each have a track: how partnering works →
03 · Why social impact
Fairness is an engineering property.
Every file above passes the same three tests: transparent enough to publish the criteria, fair enough to mark plain wording as it marks fluent wording, empirical enough to replay when someone checks.
Assessment access
A defensible grade costs hours of a marker's week, so where markers are scarce, qualifications are scarce. The Labs work on making the grade cheap enough to offer where it has never been offered, without loosening what it takes to earn one.
Arabic as a first-class language
Most systems mark Arabic writing through translation, and translation flattens the evidence. The Labs treat Arabic as a marking language in its own right: extracted, quoted and banded in the words the learner actually wrote.
Public-interest casework
When a public body refuses a licence or a claim, the person refused should be able to read why. Casework judged against published criteria carries its full rule trail, so a refusal can be checked line by line, not just appealed on faith.
The standards behind these files are written down: our commitments →
Bring a problem worth a file.
A Labs programme opens the way access does: the problem stated, the criteria written down, the status on record from the first week.
It starts with a 30-minute briefing: the problem scoped, the criteria named, a file opened or honestly declined.