A five-week online program for engineers who want to deploy AI inside a client’s business. Follow one project from the first scoping call to the moment another team can operate it.
Become the engineer a client trusts to deploy AI.
Change what you can show in your next interview. Walk a hiring manager through a decision you owned, a failure you recovered from and a result the client could measure.
01Engineer
Describe a client workflow and establish its baseline.
02Field-ready builder
Connect the system to real constraints and prove it works.
03FDE Gulf Certified
Earn nine Skill Cards and pass the capstone review.
04Your next role
Bring a redacted case study and a verifiable credential to interviews.
The field curriculum.
January 4 – February 5, 2027 · 25 lessons · 5 live workshops
Week 1 · Gulf Field Kit (Jan 4–8)
Goal
Map the Gulf client landscape and ship a first prototype on in-region models and tools.
Progression
landscape → hosting → context → tools → prototype
Promise
Arrive at a Gulf client with a working prototype in 48 hours.
Live workshop, Thu Jan 7
MCP server over a Gulf service API.
The Gulf Deployment LandscapeBuyers · Sovereign AI · Estates · Vendor Rules · CalendarAgenda
Buyers ministries and authorities, semi-government groups, banks and insurers, energy, family conglomerates, who owns budget vs who owns IT
Sovereign AI national AI strategies, in-country compute, G42 and Core42, HUMAIN, local-model preference, what "sovereign" means in a tender
Estates Microsoft, Oracle and SAP footprints, on-prem data centres, government clouds, Arabic-first legacy applications
Vendor Rules vendor registration, Etimad, local partner, in-country value and local-content requirements
Calendar Sunday–Thursday vs Monday–Friday weeks, Ramadan hours, national-day and year-end freezes, Hijri dates in systems
Models and Hosting in RegionModel Selection · In-Region Endpoints · Sovereign Cloud · Self-Hosting · Arabic Models · CostAgenda
Model Selection capability tiers, reasoning effort, context window, multimodal input, open-weight vs hosted, own-task tests over benchmarks
In-Region Endpoints hyperscaler regions in UAE, KSA, Qatar and Bahrain, model availability lag, tool calling, structured outputs, batch and caching support by region
Sovereign Cloud national and government clouds, Core42, HUMAIN, when a tender requires them, onboarding lead time
SSO OIDC, SAML, Entra ID, group claims, conditional access
RBAC roles, row- and document-level security, inheritance, least privilege
Agent Identity service principal vs on-behalf-of user, token exchange, scoped credentials, per-tool permissions
Secrets vaults, rotation, managed identities, no keys in prompts or logs
UAE Pass, Nafath, Tawtheeq national digital identity, login and signing flows, citizen- and resident-facing services, sandbox access, when an agent may act for a user
Enterprise Systems and Legacy DataCore Systems · APIs · Databases · Files · SyncAgenda
Core Systems SAP OData and BAPI, Oracle EBS and Fusion REST, Dynamics 365, ServiceNow, core banking and insurance platforms
Cross-Lingual Search multilingual embeddings, BM25 with Arabic normalization, hybrid, query rewriting, Arabic question over English sources and the reverse
Runbooks symptoms, checks, fixes, contacts, tested by a second engineer
Triage model, tool, data or infrastructure fault, reading a trace, reproducing from logs
Provider Events outages, rate-limit changes, model deprecations and migrations, re-running evals
Postmortem timeline, root cause, new eval case, added guardrail
Hypercare first-week support, issue queue, fix vs change request
HandoverCode and Infra Transfer · Documentation · Local Team Training · Outcome · Reusable KitAgenda
Code and Infra Transfer repos, IaC, pipelines, secrets ownership, licences
Documentation architecture, data flows, eval suite, runbook, known limits
Local Team Training client engineers run a release and a rollback unaided, operators handle exceptions, Arabic operator guides, national-talent upskilling requirements
Outcome baseline vs live, sample sizes, what the numbers do not prove
Reusable Kit templates, MCP servers, eval harness, lessons back to the product team
Becoming an FDE in the GulfDecomposition Interview · System Design · Live Build · Portfolio · Regional EmployersAgenda
Decomposition Interview vague client problem to plan, clarifying questions, trade-offs out loud
System Design agent in a regulated client environment, residency, failure modes
Live Build timed task on unfamiliar data and APIs
Portfolio capstone case study, repo, redaction, Skill Cards
Regional Employers labs and hyperscalers, sovereign-AI companies, integrators and consultancies, in-house teams at banks and government entities
01+Engineers who already write production code and want to deploy AI inside client organizations in the Gulf.
02+Engineers who use coding agents such as Claude Code, Codex or Cursor and want to own delivery past the prototype.
03+Practitioners who want feedback on a running system and its handover from people who deploy for a living.
04−Beginners looking for an introduction to coding.
05−Anyone looking for an introductory AI course or pure theory.
06−Anyone who needs a certificate based on attendance.
The journey · 9 Skill Cards · 1 certificate
Engineer → FDE
After FDE Gulf you become the engineer people trust when AI has to work inside real client systems, under Gulf data rules and on a real deadline. Nine Skill Cards prove each step, and one certificate shows you can take a deployment from the first conversation to handover.
⌖SC / 01 · Client Scoping
Turn a vague client request into a scoped, testable brief
Map the process, measure a baseline (cycle time, error rate, cost per case) and inventory the data you can access
Write a one-page brief with a named owner and an acceptance test on a labelled sample
◎SC / 02 · Gulf Systems Context
Map the data, hosting and identity constraints of a Gulf deployment
Classify data under UAE and KSA PDPL and Qatar’s PDPPL, and choose in-region endpoints or sovereign cloud
Plan UAE Pass or Nafath sign-in and record every decision in a residency diagram with named sign-offs
◇SC / 03 · System Design
Design an LLM system the client can run, afford and secure
Choose models, RAG or long context, MCP tool calls and structured outputs, with a token cost per task
Record each trade-off in an architecture decision and build a thin end-to-end slice first
≋SC / 04 · Retrieval Quality
Make every answer traceable to the client’s own sources
Run Arabic OCR and normalization, then hybrid BM25 and embedding search with a reranker
Measure recall@k and citation precision on a golden set, and trace bad answers to their chunk
ΔSC / 05 · Production Evals
Prove each release beats the baseline before it ships
Build eval datasets, Arabic and English LLM-as-judge rubrics and agent trajectory checks
Gate every release in CI with a threshold for each data slice
↻SC / 06 · Agent Reliability
Keep agent workflows correct when tools, models and APIs fail
Handle state, timeouts, rate limits, idempotency keys and prompt injection in tool results
Prove recovery with bounded retries, fallbacks, human escalation and failure injection
⇄SC / 07 · Deployment Safety
Ship to the client tenant without risking their operations
Package with containers and IaC, and run CI/CD with managed secrets behind a model gateway
Release through shadow mode, a canary and an approval gate, with a rehearsed rollback
∿SC / 08 · Observability
Trace latency, cost and quality for every request
Instrument OpenTelemetry traces with token cost, quality signals and prompt and model versions
Set SLOs and alerts, and debug any failed run end to end from a single trace
≡SC / 09 · Handover & Measurement
Hand over a system the client’s team can operate and extend
Transfer code and infrastructure with runbooks and an on-call triage path
Train operators to lead an incident drill alone, and measure the outcome against the week-one baseline
FDE
Credential preview
FDE Gulf Certified
After all nine cards are reviewed and your capstone passes, you receive the FDE Gulf certificate and a public verification link.
After certification you can join the FDE Gulf Talent Pool. Companies in the UAE and KSA can use it to find engineers who already understand local constraints. We make the introduction.
Jakub has 20+ years building and integrating systems. He focuses on context, tool use and human approval.
Marcin
Marcin has 15 years with SAP and complex workflows. He connects models to real client APIs.
Marcin
Marcin prototypes AI applications and tests them against the tasks people actually do. He has used low-code and no-code tools to try out product ideas before committing to a full build. His work on digital claims and payments informs how he chooses test cases and measures whether a prototype is worth deploying.
Leszek
Leszek has 25+ years in technology. He works on knowledge agents and retrieval that stays grounded.
Krzysztof
Krzysztof has 20+ years in delivery and operations. He owns use-case selection and secure deployment.
Never miss a drop.
Field notes, new resources and cohort announcements.
Tools for field work.
Vercel
OpenAI
Anthropic
Microsoft Copilot
Hugging Face
Docker
LangChain
Build a tool for one client workflow. Take the code, tests and runbook into your next deployment.
FAQ
Have another question? Ask the program team.
About the program
Who is this for?
Engineers who already ship code and want to do client-facing AI deployments in the Gulf.
Can I apply from outside the Gulf?
Yes. The projects focus on the UAE, Saudi Arabia, Qatar and Egypt. You can register your interest from another country.
What should I already know?
You should be comfortable writing code, calling APIs and debugging an application. Bring a project you can discuss without sharing confidential data.
Can I use my current engineering stack?
Yes. The reviews look at system behavior, the quality of your evidence and whether another engineer can operate your release. Document the stack you choose.
Where can I compare compensation?
The salary benchmark in the public Salary & Role Guide lists FDE Gulf estimates for Dubai, Doha, Abu Dhabi and Riyadh, with an independent UAE comparator. The guide also explains how to compare allowances and travel obligations. Read the salary benchmark →
Format and time
When does the first cohort start?
The first cohort starts January 4, 2027. Live workshop times will be shared before enrollment opens.
Which time zone will the schedule use?
The final schedule will show Dubai time (UTC+4) and Riyadh/Doha time (UTC+3). Dates and times will be published before paid enrollment.
What happens in a live workshop?
You bring that week’s artifact to a working session. Each lab includes a decision or failure to handle and a concrete acceptance check. The five workshops are listed in the agenda.
What should I bring to the first session?
Bring a workflow you can describe, one example input and a way to measure success. An anonymized client scenario or a project of your own both work.
Do I need access to client production systems?
No. Use an environment you control and data you have permission to use. A reproducible exercise with explicit limits is suitable for the review.
Skill Cards and certification
How do I earn the certificate?
Submit evidence for all nine Skill Cards and pass a practical capstone review. Attendance alone does not earn the credential.
Can I get a credential without joining the cohort?
Yes. Experienced engineers can take the free assessment with evidence from prior work. If all nine Skill Cards pass, you earn FDE Gulf Assessed. Get assessed for free →
What is the difference between the Certificate and “Assessed”?
The full Certificate includes live workshops, full Skill Card reviews and priority in the Talent Pool. Assessed is for people who already have the work and pass the free assessment.
Does the starter card count toward certification?
The starter card is a private self-assessment. The credential requires nine reviewed Skill Cards and a successful capstone review.
Can employers verify my credential?
Yes. Issued certificates have an individual ID and a public verification page that shows their current status.
Do I get job introductions?
Yes. Certified participants get priority introductions through the Talent Pool. See how hiring works →
What can I put in my interview portfolio?
Use your redacted brief, architecture decision, evaluation report, release drill and capstone. Obtain permission before sharing work that belongs to a client.
Pricing and enrollment
What does it cost?
Early bird: 5,499 AED until December 14, 2026. Regular price: 6,999 AED. Joining the waitlist is free, and so is the independent assessment.
How long do I have access to materials?
Paid cohort participants have 12 months of access to all materials from the cohort start date.
What if I’m not strong enough?
After purchase you get a short readiness check. If it’s clearly not a fit, you can get a refund within 7 days or move to a later cohort.
Pick one system you’ve built. Turn its strongest evidence, its weakest assumption and your next experiment into a card you can take to an interview.
FIELD CARD challenge.md
Read https://fdegulf.dev/challenge.md. Interview me about one deployment and help me create my first FDE field card. Use only details I choose to share. Keep the card local.
Paste into your coding assistant, or read the instructions and complete it yourself.
Become the FDE people trustwith production AI.
Early bird 5,499 AED ends December 14. 25 seats in the first cohort.