Founder-led AI consultancy · Dubai

AI consulting for document-heavy, regulated organizations.

Signet helps UAE organizations assess, build and govern document AI, knowledge assistants and workflow automation, with human review designed in from the start.

Book a consultation Explore our work
  • ForGovernment, healthcare, insurance, banking and real estate teams in the UAE
  • SolvesSlow manual document review, knowledge that's hard to find, and AI vendor choices made without evidence
  • DeliversAn opportunity assessment, a pilot on your own data, or production delivery with governance
Illustrative example. Invented data, not a live model.
Motor claim note, received by email

Policyholder reports a rear collision on 14 March 2026 at Sheikh Zayed Road, Dubai. Policy MTR-20418-77. Garage estimate AED 8,450. Other driver may have been at fault, police report to follow.

Structured recordIllustrative confidence
incident_date2026-03-140.98
locationSheikh Zayed Rd0.95
policy_noMTR-20418-770.99
estimate_aed84500.97
liabilitythird party?0.61
How a document AI workflow treats uncertainty: confident fields flow through, and low-confidence fields go to a person before reaching your systems.

Professional experience

  • Dubai Health AuthorityAI Specialist · employment
  • SanlamLead ML Engineer · employment
  • Nomo FintechData Scientist · employment
  • DigsFactML Engineer · employment and contract
  • OmdenaML Engineer · contract projects
  • DetectNowML Engineer · research project

Organizations where Mohamed has worked or contracted before founding Signet. They aren't presented as Signet clients.

Credibility at a glance

  • More than six years delivering applied AI and ML
  • Seven AWS, Google and Microsoft certifications
  • MongoDB User Group leader in Dubai and Cairo
Engagements

Three ways to work together

Each engagement stands on its own, and each one ends with a clear recommendation, so you only commit to the next stage when the evidence supports it.

Supporting capabilities

Applied across all three engagements.

  • Document AI and extractionForms, contracts and records turned into structured, reviewable data
  • Knowledge assistants (RAG)Answers grounded in your documents, with sources cited
  • ML engineering and computer visionClassification, prediction, detection and segmentation on AWS
  • Requirements and vendor evaluationRFP requirements, scoring criteria and evidence-based testing
  • AI governanceSelection, testing, privacy and human oversight processes
  • ToolsAWS Bedrock, SageMaker, Lambda, Azure AI, Python, FastAPI, LangChain, MongoDB
Case studies

Selected work

Delivered by Mohamed in employed, contract and research roles before founding Signet. Each case states the relationship, his responsibilities and how far the work went. Nothing here is presented as a Signet client engagement.

Healthcare Employment

Enterprise AI requirements for patient record summarization

Dubai Health Authority · AI Specialist · Sep 2025 to Aug 2026

Business problem

A large public health authority wanted to adopt AI for clinical and operational work, including summarizing patient records across specialties, without compromising safety, privacy or continuity of care.

Mohamed's responsibilities

  • Technical scoping and evaluation of enterprise AI initiatives
  • Translating healthcare needs into RFP requirements, solution criteria and vendor assessments
  • Coordinating business, IT and vendor teams; developing AI governance and repeatable selection processes

Key decisions

Specialty-specific summary outputs instead of a single generic format. Confidence thresholds paired with clinician review. Privacy, integration, service levels and disaster recovery written in as requirements rather than left to vendors.

Outcome

Requirements and evaluation criteria were defined for the summarization initiative. Procurement and deployment results aren't reported here.

Sample deliverable: the seven requirement areas defined for patient record summarization Requirements specification · Patient record summarization Specialty-specific outputsConfidence thresholdsClinician review stepPrivacy and data handlingIntegration with records systemsService levelsDisaster recovery DefinedDefinedDefinedDefinedDefinedDefined Human gate
Sanitized illustration of the requirement areas listed above, not the original document.
Stage
Requirements and vendor evaluation
Relationship
Employed role
Focus
Requirements, governance, vendor assessment
Insurance Employment

Knowledge retrieval across about 60,000 pages of product content

Sanlam · Lead Machine Learning Engineer · Oct 2023 to Jan 2025

Business problem

Insurance product information was spread across documentation from several subsidiaries, making consistent answers about products hard to find.

Mohamed's responsibilities

  • Led ML and LLM solutions for insurance workflows, from data collection and feature engineering to model integration
  • Built retrieval over roughly 60,000 pages of product content
  • Developed document extraction workflows on AWS Bedrock and Claude

Key decisions

Retrieval augmented generation over the source documentation, so answers are grounded in the actual product content. Extraction workflows use application tools to convert content into structured records rather than free text.

Outcome

Solutions were integrated into preproduction systems. Usage and accuracy metrics aren't reported here.

Retrieval pipeline: product documents are indexed, relevant passages retrieved, and answers returned with their sources ~60,000 pagesseveral subsidiaries Indexed passages Is X covered abroad? source: p. 214 Answer with cited source Schematic. Question, answer and page number are illustrative.
Retrieval augmented generation: answers grounded in, and citing, the source documents.
Stage
Preproduction
Relationship
Employed role
Scale
About 60,000 pages
Insurance and materials Contract Employment

Computer vision for material and insurance damage analysis

DigsFact Inc. · ML Engineer · contract Jun to Aug 2022, then employment Sep 2022 to Jul 2023

Business problem

Assessing material condition and insurance damage from images is slow and inconsistent when done by eye.

Mohamed's responsibilities

  • As a contractor, delivered two client proofs of concept for material and damage detection
  • As an employee, built image segmentation and object detection models
  • Hosted the models on AWS

Key decisions

Segmentation for the extent of damage combined with object detection for locating it, validated first through client proofs of concept before further model development.

Outcome

Two client proofs of concept delivered, followed by eight models hosted on AWS.

Schematic of damage detection: a vehicle image with a detected damage region and segmentation mask damage region 8 modelshosted on AWS Schematic. Not a client image.
Detection locates damage; segmentation outlines its extent.
Stage
Proof of concept, then hosted models
Relationship
Contract, then employment
Scale
2 client proofs of concept, 8 models

More work

Further projects from Mohamed's employed, contract and research roles.

  • Nomo FintechData Scientist · 2024 to 2025

    LLM document extraction with AWS Bedrock and Claude, turning unstructured documents into structured attributes; automated ML pipelines on AWS; retrieval over internal Confluence content.

    Employment
  • Dubai Health AuthorityAI Specialist · 2025 to 2026

    Supported design and evaluation of HR agents for job descriptions, interview questions and screening, with a pilot and human oversight. Scoped an AI-assisted sick leave audit workflow and contributed to fraud, waste and abuse solution redesign.

    Employment
  • OmdenaML Engineer · 2021 to 2023

    Four client projects covering image segmentation, satellite image super resolution, analytics, and manufacturing automation including damage detection on a production line.

    Contract
  • OmdenaJunior ML Engineer · 2020

    Alternative-data credit scoring using correlation analysis for applicants without traditional credit histories.

    Contract
  • DetectNowML Engineer · 2020 to 2021

    Cough sound classifier supporting lung disease research, with clinical data collection and audio cleaning alongside medical colleagues. The project was covered by Forbes and CNNMoney Switzerland in 2020.

    Research
  • Freelance and internship2019 to 2022

    Deployed a healthcare image retrieval model and built a text classifier exposed through a REST API.

    Contract
Delivery process

How an engagement runs

  1. Introductory call

    We discuss the workflow, the documents involved and your constraints, and decide whether Signet is the right fit.

  2. Written scope

    A short proposal: the engagement type, success measures, what you provide, what you receive and how data is handled.

  3. Delivery with checkpoints

    Regular reviews with your process owners. Nothing reaches your systems or users without your sign-off.

  4. Recommendation and handover

    Findings, documentation and a clear owner on your side, plus a recommendation for what comes next.

Human review by design Data handling agreed in writing first Success measures set before building Your team owns the result
Portrait of Mohamed Amr, founder of Signet
Mohamed AmrFounder, Signet · Dubai
Founder-led

You work directly with Mohamed

Signet is the consultancy of Mohamed Amr, a senior AI and ML engineer in Dubai. You deal with him from the first conversation to handover.

For more than six years Mohamed has built machine learning, document AI, retrieval and computer vision systems, and led enterprise AI requirements, vendor evaluation and governance in healthcare, insurance and fintech.

He works on both sides of the table: writing Python and deploying on AWS, and sitting with clinicians, underwriters and IT leads to decide what should be automated and what needs a person in the loop.

  • Stakeholder discoveryTurning business needs into clear requirements
  • Vendor evaluationRFP criteria and testing on representative data
  • Hands-on implementationPython, AWS Bedrock and SageMaker, APIs
  • HandoverDocumentation and ownership your team can run with

Certifications

  • Machine Learning, SpecialtyAWS
  • Data Analytics, SpecialtyAWS
  • Database, SpecialtyAWS
  • Solutions Architect, AssociateAWS
  • Cloud PractitionerAWS
  • TensorFlow Developer CertificateGoogle
  • Azure AI FundamentalsMicrosoft

Education

  • MSc Data Science In progressUniversity of Leeds · thesis pending
  • BSc Computer and Systems EngineeringAin Shams University · 2023

Community and recognition

Insights

Notes from practice

Practical writing on document AI, retrieval and AI governance.

Earlier posts on MediumAll articles
FAQ

Common questions

Is Signet a team or one person?

Signet is founder-led. You work directly with Mohamed Amr from the first conversation through delivery and handover.

Which organizations do you work with?

Organizations in the UAE where documents are dense, data is sensitive and decisions need human review, such as government entities, healthcare providers, insurers, banks and fintechs, and real estate organizations.

Where should we start if we have no defined use case?

With an AI Opportunity Assessment. It compares candidate workflows on value, data readiness, integration effort and risk, and ends with a prioritized shortlist and a roadmap.

Can you help us evaluate AI vendors?

Yes. Mohamed writes requirements and evaluation criteria and tests vendor claims against representative documents, including security, privacy and service levels.

How is confidential data handled?

Data handling is agreed in writing before any work starts, including what data is shared, where it is processed and who can access it. Anonymized samples are used wherever they are enough.

Do you work outside Dubai?

Signet is based in Dubai and works across the UAE. Remote engagements elsewhere can be discussed.

Book a consultation

Tell us about the workflow you want to improve.

The first conversation is a chance to see whether AI fits, and which engagement makes sense.

  • The workflow, the documents involved and where time is lost today
  • Constraints: data sensitivity, systems, regulation and approvals
  • Whether an assessment, a pilot or a vendor review is the right first step

Request a consultation

Fill this in and your email app opens a ready-to-send message. Prefer not to? Email directly or message on LinkedIn.

Please don't include confidential or personal data here.

This page doesn't store or send your details. They're placed into an email that you review and send yourself. See the privacy notice.