AI Data & Human Intelligence · Arabic · RLHF · Domain Expertise

The AI training infrastructure the world's frontier labs depend on.

Expert human workforces. Sourced, vetted, and managed on-site. Delivering Arabic annotation, RLHF, and domain expertise to AI labs and enterprises across the USA, Europe, and the Gulf.

Work with us → See capabilities
0+
Expert professionals in pipeline
0 yrs
Operational track record
0
Markets served · USA · EU · Gulf
Delaware
Incorporated · USA
Built for

Expert data for
frontier AI labs.

Frontier AI labs · Enterprise AI teams · Gulf AI programmes · Multilingual NLP teams

RLHF

Preference ranking at scale

Pre-vetted domain experts delivering RLHF, red-teaming, and model evaluation to US and EU labs — to frontier standards.

Annotation

Multilingual training data

Arabic, English, and multilingual annotation pipelines. Managed on-site. Auditable, high-security, delivery-tracked to your specification.

Domain expertise

Medicine · Law · Finance

Specialist annotators with verified credentials in high-stakes domains — for the training tasks general crowdsourcing cannot handle.

On-site managed

Controlled environments

El Gouna facilities, with contractors housed and managed. Auditable infrastructure — the premium Scale and Surge charge for, built in MENA.

Capabilities

Three capabilities.
One operating machine.

One operational model, proven across five years: sourcing, vetting, housing, and managing expert human workforces to deliver AI training data at the standard frontier labs demand.

I

Arabic & Multilingual Data Annotation

Native Arabic dialect annotation at the cultural register and linguistic precision no remote platform can replicate — Modern Standard Arabic through Gulf, Levantine, and Egyptian vernacular, with the context that training data for 400M+ speakers demands.

  • MSA, Gulf, Levantine, and Egyptian dialect annotation
  • Cultural context and register calibration — not translation
  • Sentiment, instruction-following, and preference data in Arabic
  • Multilingual parallel corpora for cross-lingual training
Arabic NLPDialectRegisterInstruction tuning
Discuss this →
II

RLHF & Alignment Contractor Pools

Pre-vetted domain experts in medicine, law, engineering, and finance — deployed to US and European AI labs for preference ranking, model evaluation, red-teaming, and alignment work, to the bar frontier development demands.

  • Preference ranking and comparative evaluation
  • Adversarial red-teaming and safety evaluation
  • Domain-expert labelling: medical, legal, financial, technical
  • Scalable output under NDA and data-governance frameworks
RLHFPreferenceRed-teamingSafety eval
Discuss this →
III

On-site Managed Operations

Contractors housed and managed on-site in El Gouna — a master-planned, infrastructure-complete Red Sea community. Controlled, auditable, high-security annotation environments: the infrastructure premium Scale AI and Surge AI charge for, built in MENA five years ago.

  • Housed contractor workforce in a controlled facility
  • Auditable delivery environment — contractually documentable
  • NDA, IP protection, and data-security protocol compliance
  • Payroll, compliance, and HR managed in-house
On-siteAuditableNDA-readyIP-secure
Discuss this →
What you get

Data your models can actually train on.

Not a vendor. An infrastructure partner with skin in the delivery.

Training data that passes your quality bar

Inter-annotator agreement tracked. Delivery auditable. Every batch reviewable before it enters your pipeline.

Arabic at dialect depth, not translation quality

Gulf, Levantine, Egyptian, and MSA — natively. Not translated, not generic. The register your model needs to think in.

Domain experts, not crowdsourced generalists

Medical, legal, financial, and engineering tasks go to verified professionals. The credential is checked, not assumed.

You receive data. Not a management problem.

Kalasko runs quality control, inter-annotator agreement, and delivery cadence. Your team reviews output, not people.

Quality

The quality bar is the product.

Frontier training data is only useful if you can trust every label. Quality isn't a final check at Kalasko — it is built into how the work is run.

01

Inter-annotator agreement, tracked

Every batch is measured for agreement. Disagreement is adjudicated and resolved — not shipped.

02

Calibrated against gold sets

Annotators are measured against reference answers before a project starts and throughout, so the standard holds at scale.

03

Reviewed before delivery

Nothing reaches your pipeline unreviewed. Every batch is auditable and reviewable before you accept it.

04

Verified experts, consistent reviewers

Credential-checked specialists, managed on-site — the same reviewers across a project, never an anonymous crowd.

Structural advantage

Why Kalasko is
structurally
unreplicable.

Three barriers keep any competitor — however well-resourced — from entering this space quickly. They are not technology barriers. They are operational, linguistic, and legal barriers, built over five years from the inside out.

Replicating Kalasko from scratch takes years before first reliable delivery. Working with Kalasko takes a conversation.

01

Arabic dialect depth

400M+ speakers. Dozens of dialects. A cultural register impossible to replicate through translation or remote crowdsourcing. Egyptian talent natively reads Gulf, Levantine, and MSA — at once.

Linguistic moat
02

Physical on-site infrastructure

El Gouna gives what no remote platform can: a stable, master-planned community with housed contractors, reliable utilities, and auditable physical delivery — built in MENA first.

Operational moat
03

Delaware + Egypt dual structure

Incorporated in Delaware — the standard US and European counterparties expect — while operating in Egypt. Five years of local regulatory, banking, and employment compliance, built in.

Legal moat
04

Five years of operational proof

600+ professionals through the pipeline. Active clients across the USA, Europe, and the Gulf. The track record exists. The machine has been running since 2020.

Track record

Build from scratch

  • ✗12–24 months to source and vet a local talent base in a new geography
  • ✗Housing, payroll, and compliance for on-site contractor management
  • ✗Regional client relationships and an auditable delivery record, from zero
  • ✗Egyptian regulatory, banking, and employment law without local knowledge

Realistic: two to four years before first reliable delivery.

Work with Kalasko — day one

  • ✓600+ professionals already sourced, vetted, and in the delivery pipeline
  • ✓On-site housing and managed operations in El Gouna, running since 2020
  • ✓Active client base across the USA, Europe, and the Gulf — proven delivery
  • ✓Arabic-native operational leadership with deep Gulf and Egypt experience

A Delaware Inc. structure, clean for US and EU counterparties, from day one.

Underserved language · Global AI gap

The world's biggest
untapped AI data
opportunity.

Arabic has 400M+ native speakers, is the official language of 26 countries, and sits at the centre of major Gulf AI initiatives. Yet it remains critically underserved in global AI training data — not for lack of demand, but because the supply chain needs capabilities remote-first platforms have not built.

Kalasko sits at the intersection of Egyptian STEM talent, Arabic-native operational leadership, and on-site managed delivery. No other platform combines all three.

400M+
Native Arabic speakers
26
Arabic-speaking countries
4
Dialect families covered
Dialect coverage · Kalasko annotators
Gulf Arabic
Saudi · UAE · Kuwait · Qatar · Bahrain · Oman
Native coverage
Egyptian Arabic
Most widely understood dialect across MENA
Native base
Levantine Arabic
Lebanon · Syria · Jordan · Palestine
Annotator pool
Modern Standard Arabic
Formal · cross-regional · media and legal register
Full coverage
Discuss Arabic data needs →
How it works

From brief to
delivery, in weeks.

No procurement cycle. Kalasko's operational infrastructure is built to move from first conversation to first delivery inside two to four weeks for most AI data tasks.

  1. 1

    Scope the data task

    A direct conversation: what you need, at what volume, to what standard. Arabic annotation, RLHF pools, domain-expert labelling, or on-site managed work — translated into a spec our annotators can execute immediately.

    Annotation specVolume & timelineQuality barLanguage requirements
    1–2 business days
  2. 2

    Talent matching & vetting

    The task is matched to the right pool — by domain expertise, language register, and verified credentials. Domain-expert tasks go to annotators with verified professional backgrounds.

    Domain matchingCredential verificationDialect confirmation
    2–5 business days
  3. 3

    Legal & data governance setup

    A Delaware incorporation means US and European clients face zero legal friction. NDA, IP assignment, and data-governance frameworks are standard — not bespoke. We operate under your security protocols or bring our own auditable framework.

    NDA & IP assignmentData governanceDelaware MSA
    Parallel to matching
  4. 4

    Managed delivery

    Annotators work on-site in El Gouna under direct management. Quality control, inter-annotator agreement, and delivery cadence are run by Kalasko — not delegated to the client. You receive data, not a labour-management problem.

    On-site managementQC & IAA trackingDelivery cadence
    Ongoing from week 2–4
  5. 5

    Scale on demand

    Egypt produces one of the largest STEM graduate pools in the region. Kalasko's sourcing pipeline is continuously active. When a project scales, the talent base is already there — no six-month recruiting cycle, no new infrastructure.

    Elastic poolContinuous sourcingNo ramp-up
    On request
Operational scale

Running at scale,
today.

0+
Active contractors managed today
0+
Contractors housed on-site
2–4 wks
Brief to first delivery

Every number above is verifiable. Operational references available under NDA on request.

The next phase of AI will be shaped by the languages it can think in. Arabic is the most underserved major language in global AI training data. The supply chain to fix that has been running in one place since 2020.
Kalasko · 2026

Active clients across
three markets.

USA, Europe, and the Gulf. The client base validates the delivery model and the Delaware structure across all three vectors at once.

TreviPay
United States
B2B Payments
Qreach
United Kingdom
Technology
Halo Platforms
United States
AI · Technology
Guretruck
Europe
Logistics
Contactless.io
Europe
FinTech
MDC B.V.
Netherlands
Media
Westward →

USA & Europe

US and European AI labs sourcing annotation, RLHF, and domain-expert capacity. A Delaware structure removes the legal friction. Clients across payments, logistics, media, and technology validate the model.

Live · Scaling
Eastward →

Gulf & MENA

Saudi Vision 2030, the UAE AI strategy, NEOM, and SDAIA are creating unprecedented demand. Arabic-native leadership and an Egyptian talent base position Kalasko as the natural MENA partner.

Active · First mover
Infrastructure ◆

Operational base

On-site facilities with housed contractor workforce, auditable delivery environments, and access to one of the largest STEM graduate pools in Africa and the Middle East. Operational since 2020.

Operational since 2020
From the field

What working with Kalasko looks like.

Brief to first delivery batch in two to four weeks — no procurement cycle, no six-month ramp.

Typical engagement · brief to first batch

Quality control, inter-annotator agreement, and delivery cadence are Kalasko's to run. You review output, not process.

On-site managed model

Delaware MSA and NDA in place on day one. First annotators matched by day three. No legal friction.

Onboarding · Delaware structure

Representative of how Kalasko engagements run. Specifics vary by project. References available under NDA on request.

Founded by

Two founders.
One rare combination.

AI depth and regional execution: academic rigour, decades of enterprise technology, and native Arabic operational fluency across the markets Kalasko serves.

S

Sameh (Sam) Abdalla

Co-Founder
  • PhD Computer Science · Università di Trento
  • MSc Management · Università Cattolica, Milan
  • CTO · Intouch.com · Europe & Latin America

Senior research at Siemens Milan, postdoctoral fellowships at University College Dublin and the University of Lille, and over a decade as CTO of Intouch.com — an AI-powered retail media platform across Europe and Latin America. Deep technical credibility in AI and data systems, paired with hands-on experience scaling technology in international markets.

AI & data systemsCTO · product scalingEU & LatAmPhD Computer Science
A

Ahmed Serag

Co-Founder & Managing Director
  • CS · University of Nottingham
  • National Bank of Abu Dhabi · I-SCORE · Orange
  • Arabic–English native bilingual

15+ years in enterprise technology across the Gulf and Egypt — National Bank of Abu Dhabi, the Egyptian Credit Bureau (I-SCORE), the Egyptian Collateral Registry, and Orange Business Services. Running Kalasko's operations since founding in 2020. Native bilingual in Arabic and English — an irreplaceable asset across the Western delivery model and the Gulf expansion.

15 yrs enterprise techGulf & Egypt opsArabic · English nativeNetwork security
Questions

Answers, before you ask.

What exactly does Kalasko deliver?
Arabic and multilingual data annotation, RLHF and alignment contractor pools, and on-site managed domain-expert labelling — for AI labs and enterprises across the USA, Europe, and the Gulf.
How fast can you start — and deliver?
The first conversation is with the founding team, not an SDR. Most engagements move from brief to first delivery batch in two to four weeks. No procurement cycle, no six-month ramp.
How do you protect our data and IP?
A mutual NDA and full IP assignment are in place before any data moves. Work runs in a controlled, on-site facility under your security protocols or our own auditable framework — see security & data governance.
How do you guarantee annotation quality?
Credential-verified experts, inter-annotator agreement tracked on every batch, calibration against gold sets, and a review pass before anything enters your pipeline.
Which languages and dialects do you cover?
Native Arabic across Gulf, Levantine, Egyptian, and Modern Standard registers — at cultural and dialectal depth — plus English and multilingual parallel data.
How do we engage?
Email sam@kalasko.com or use the form below. Every enquiry is handled directly by the founding team, and a full NDA is available before any data is shared.
Security & data governance

Your data never leaves a controlled environment.

The biggest risk in remote data work is leakage — to home machines, anonymous crowds, and untraceable hands. Kalasko removes it by design.

Work runs on-site under direct management, under your security protocols or our own auditable framework. Nothing is improvised.

01

On-site controlled environment

Annotation happens in a managed, housed facility — not on anonymous remote machines. Physical control over who touches your data.

Physical control
02

NDA & IP, from day one

A mutual NDA and full IP assignment are standard, signed before any data moves. Your data and your outputs stay yours.

Contractual
03

Auditable, documentable delivery

Controlled delivery environments with a traceable record. Every batch is reviewable and contractually documentable.

Auditable
04

Delaware legal structure

A US entity incorporated in Delaware gives US and European counterparties clean, familiar legal recourse.

Legal
Already running.

The operational model behind the names defining AI data — domain-expert vetting, on-site management, auditable delivery — Kalasko has run since 2020.

They are still building the infrastructure. Kalasko already operates it.

None of them serve the Arabic market. None operate on-site in MENA. The window to work with the platform that already does — before it scales further — is open now.

Work with us → Or email sam@kalasko.com directly
Work with us

Arabic annotation.
RLHF capacity.
Managed domain
experts.

Tell us what you need. We respond within one business day. Every enquiry is handled directly by the founding team — no SDRs, no qualification flows.

  • US and European AI labs sourcing annotation and RLHF capacity
  • Gulf AI programmes needing Arabic-native data operations
  • Enterprise AI teams requiring domain-expert labelling at scale
  • Platforms exploring MENA data supply-chain access
Or reach us directly
sam@kalasko.com

We respond within one business day. Full NDA available on request before any data is shared.

We'll be in touch.