Registered in Meydan Free Zone, Dubai

AI systems and human data operations built for real-world deployment.

ABI AI builds custom AI products for organisations, and helps AI companies design and operate the high-quality human-data programmes their models depend on.

How human data becomes model improvement

The human-data feedback loopDomain experts feed into task design and rubrics, which produce evaluation and preference data, which surfaces model behaviour — regressions and capability gaps. Findings feed back to the experts, closing the loop.Domain expertsRecruited, calibrated, and measuredTask design & rubricsSpecification, edge cases, worked examplesEvaluation & preference dataGold sets, agreement, audit samplingModel behaviourRegressions and capability gaps identifiedFEEDBACK LOOP
Legal entity
ABI AI L.L.C-FZ
Jurisdiction
Meydan Free Zone, Dubai
Established
2026
Current customer
G2i

What we do

Two disciplines that support each other.

Building AI systems well and producing good human data are the same problem seen from two directions. We work across both.

AI Product Development

We design and build AI systems that go into production — not prototypes that stall after a demo.

  • Custom AI applications and internal tools
  • LLM-powered assistants and agent workflows
  • AI-enabled workflow automation
  • Data and decision-support systems
  • Model integration and evaluation
  • Applied AI research and consulting
See how we work

Human Data Operations

We help AI organisations design and run the human-data programmes their models depend on.

  • Human data collection and expert domain data
  • Annotation, curation, and quality assurance
  • Model evaluation and preference data
  • Red-teaming and safety evaluation
  • Contributor workforce operations
  • Multilingual and specialist programmes
See how we work

Why ABI AI

Small, senior, and specific about what we do.

Senior-led engagements

Work is led directly by the founder rather than handed to a junior delivery layer. You deal with the person doing the thinking.

Research-grounded practice

Our approach comes out of peer-reviewed machine-learning research in finance, healthcare, and autonomous systems — not vendor marketing.

Operational, not theoretical

Human-data programmes are designed around throughput, quality measurement, and the people doing the work.

Transparent by default

We are a young company and we say so. Scope, capacity, and limitations are stated plainly before an engagement starts.

Current customer

ABI AI currently supports G2i’s human-data operations, with founder Soroush Farghadani serving as Head of Human Data.

This is our current customer engagement. We would rather name one real customer than imply a roster we do not have.

Founder experience

Soroush Farghadani

ABI AI was founded by Soroush Farghadani, who leads human-data work at xAI, previously led AI for autonomous drone navigation at SOTI’s advanced aerospace research group, and has published peer-reviewed machine-learning research spanning financial AI, explainable AI for healthcare, computer vision, and autonomous systems.

The experience described here is the founder’s own professional and academic background. It is not presented as work delivered by ABI AI.

Read the full profile

Frontier AI lab

Human Data — Lead at xAI

Robotics & autonomy

Lead Data Scientist, SOTI Advanced Aerospace Research

Academic background

M.Sc. Computer Science, University of Toronto

Published research

Frontiers in AI · MICCAI MLMI · arXiv

Tell us what you are trying to build or measure.

Whether it is an AI system you need built or a human-data programme you need designed and run, the first step is a straightforward conversation about scope and fit.