01About
- monthly compliance reporting with ESRA at NEOM
- 4 wks → 1 hrmonthly compliance reporting with ESRA at NEOM
- contractors adopted it as the standard procedure
- 800contractors adopted it as the standard procedure
- eval runs passed before every agent release
- 138/138eval runs passed before every agent release
- developer onboarding after automation
- 6 → 2 wksdeveloper onboarding after automation
I design systems end to end: ingestion, data modeling, retrieval, agents, UX, and the cloud platform underneath. Most of my work sits where messy real-world data meets people who need answers from it quickly.
I spent five years at NEOM, growing from DevOps to Senior Solution Architect and advising the Deputy CEO's Office on AI and data. Before that I built software at IBM in the US, and studied computer science at UMass Amherst and KAUST. Today I build bilingual (Arabic/English) WhatsApp agents for small businesses, and I write a monthly post about what I'm learning.
NowUpdated October 2026
- BuildingWhatsApp front-desk agents for clinics, salons and law firms.
- WritingA monthly post on agents, evals and LLM engineering.
- ExploringVoice agents that handle Gulf Arabic on real phone lines.
02Experience
2026 — Present
Independent AI Engineer · AI agents for small businesses
Dammam, Saudi Arabia
Design, build and run WhatsApp front-desk agents for clinics, salons and law firms in Saudi Arabia.
- Bilingual agent tuned for Gulf Arabic: FAQs, bookings, reminders, lead follow-ups and human handoff.
- Eval-gated releases: 46 scripted conversations × 3 runs, 138/138 passing at ~$0.003 per conversation.
- Exactly-once reminder scheduler on Postgres, proven with concurrent-scheduler smoke tests.
- Claude API
- FastAPI
- PostgreSQL
- WhatsApp Cloud API
Jun 2026 — Sep 2026
Senior Systems Engineer · EPAM Systems
Al Khobar, Saudi Arabia
Led DevOps and deployment for a client's enterprise AI services, ran daily requirements sessions with the client, and developed backend features.
- Owned deployments to production on GCP and OpenShift with Docker, Terraform and ArgoCD (GitOps), including secrets, TLS and LDAP/SAML authentication.
- Developed the Speech-to-Text feature for the client's HR agent, debugging it on the client's network through to a working production deployment.
- Wrote Python connection tests early in the engagement that surfaced client-network issues before they could block deployment.
- Designed PostgreSQL schemas and data distribution, consolidated migrations, and automated their execution on service startup.
- DevOps
- OpenShift
- ArgoCD
- Terraform
- Speech-to-Text
- PostgreSQL
Jan 2024 — Jun 2026
Senior Solution Architect · NEOM
Tabuk, Saudi Arabia
Trusted technical advisor to the Deputy CEO's Office: scoped AI and data solutions and presented designs and decision-intelligence tools to senior leadership.
- Built ESRA, the app contractors use for monthly environmental submissions: reporting went from ~4 weeks to ~1 hour, adopted by all 800 contractors.
- Built EnviroGPT, a RAG chatbot that answers NEOM environment teams' questions from indexed environmental documentation.
- Built LLM and RAG solutions in Palantir Foundry and led design of an environmental data platform unifying multiple waste and compliance systems.
- Designed GCP networking and storage for enterprise data platforms and set DevOps and CI/CD standards across internal apps.
- GenAI
- RAG
- Palantir Foundry
- Ontology
- GCP
Jul 2021 — Dec 2023
DevOps Specialist → Senior DevOps Specialist · NEOM
Tabuk, Saudi Arabia
Built the delivery and service platforms NEOM's internal teams ran on, and led a team of 10 consultants.
- Delivered a centralized helpdesk platform that improved resolution times by 150%.
- Cut developer onboarding from 6 weeks to 2 and application publishing cycle times by 75%.
- Automated OCI provisioning, integrated CI/CD across teams, and ran licensing and compliance for 1,000+ users.
- CI/CD
- OCI
- Kubernetes
- Automation
- Team lead
2017 — 2019
Software Developer · IBM
Austin, TX & Durham, NC, USA
Built enterprise reporting features and mobile apps, starting as an intern and returning full time.
- Built React UI components and real-time analytics integrations for enterprise reporting.
- Automated compliance and security scans in IBM DevOps Insights, improving audit efficiency by 50%.
- As an intern, shipped IBM Connections mobile features in React Native and improved app performance by 50%.
- React
- React Native
- Analytics
- DevOps
03Selected work
Independent · clinics, salons, law firms· 2026
WhatsApp AI Front-Desk Agent
A bilingual (Arabic/English) agent that runs a business's WhatsApp number like a front desk: answers, books, reminds, follows up, and hands off to staff.
- eval runs passed
- 138/138eval runs passed
- per conversation
- ~$0.003per conversation
- median reply
- 2.8smedian reply
- unit & integration tests
- 292unit & integration tests
- Claude API
- Python
- FastAPI
- PostgreSQL
- Evals
Read case study →
NEOM
ESRA: Environmental & Social Responsibility Application
The application NEOM's contractors use to submit their monthly environmental sheets, feeding validated data straight into compliance dashboards.
- monthly compliance reporting
- 4 wks → 1 hrmonthly compliance reporting
- contractors on the standard procedure
- 800contractors on the standard procedure
- less analyst effort
- ~90%less analyst effort
- adopted by other sectors
- Templateadopted by other sectors
- Online forms
- Data pipelines
- Data quality
- KPI modeling
- Dashboards
Read case study → Internal work, shared at a high level
NEOM· 2024–2026
EnviroGPT
A chatbot for NEOM's environment teams that answers questions from the organization's environmental documentation, using retrieval-augmented generation.
- GenAI
- RAG
- Chatbot
- Palantir Foundry
Read case study → Internal work, shared at a high level
NEOM
Tender360
Automated tender intelligence: unstructured bid documents become searchable data and executive-ready summaries of key points, risks and comparisons.
- tender processing time
- Weeks → hourstender processing time
- less manual effort
- ~90%less manual effort
- RAG
- Embeddings
- Semantic search
- LLM summarization
- Dashboards
Read case study → Internal work, shared at a high level
More work
Delivery Platform: CI/CD, IaC & KubernetesNEOM
Standardized pipelines, infrastructure as code and Kubernetes so teams ship on a predictable cadence with governance built in.
Outcome: Releases went from every ~6 months to every 2 weeks, onboarding from 6 weeks to 2, publishing cycle times dropped 75%, and helpdesk resolution times improved 150%.
- CI/CD
- Infrastructure as Code
- Kubernetes
- Observability
- RBAC
NEV Environmental DashboardsNEOM
Interactive dashboards covering projects by status, compliance metrics, complaints and exceedances, regional breakdowns, and diesel generator locations.
- React
- Data visualization
- Maps
Presentation Parsing PipelineNEOM
End-to-end pipeline that extracts content from PowerPoint files, with OCR fallback, slide quality checks, and logging, productionized in Foundry.
- Python
- pytesseract
- Tesseract
- Pillow
- Palantir Foundry
CAPEX Estimator for the Line FMPalantir Foundry
Foundry application for estimating capital expenditure (CAPEX) for the Line FM.
- Palantir Foundry
Oil & Gas Pipeline GeomapConfidential energy-sector client
Geospatial application mapping pipeline routes and stations, with operational dashboards. Client and project details withheld due to confidentiality.
- Palantir Foundry
- Maps
- Dashboards
Speech-to-Text for an Enterprise HR AgentEPAM Systems (client project)
Developed the Speech-to-Text feature for a client's HR agent and owned its deployment to production on GCP and OpenShift.
- Speech-to-Text
- DevOps
- OpenShift
- ArgoCD
Open source
- whatsapp-ai-agentBilingual (Arabic/English) AI agent for WhatsApp Business: FAQs, bookings, reminders, follow-ups and human handoff. Python, FastAPI, Postgres, Claude API, eval-gated.Python
- auto-exam-grader-2016Automated exam grading pipeline: OCR (Tesseract) for short answers, OMR bubble reading for MCQs, YAML rubric engine, SQLite storage and CSV/HTML reports — via CLI or Flask web UI. Cut grading turnaround ~80%.Python
- marine-buoy-simIoT marine buoy network simulator: sensor models (temperature, pH, salinity, turbidity), LoRa/Zigbee-style mesh networking, SQLite/CSV data pipeline with anomaly detection, and a Flask dashboard + REST API.Python
- seashell-shopNacré — a bohemian-luxury, bilingual (English/Arabic RTL) storefront for handcrafted seashell hangings and shell chandeliers. Responsive static site in HTML/CSS/JS with scroll-reveal animations.HTML
04How I work
- 01
Front-load the risk
Under deadline pressure I go after the unknowns first, so surprises show up while there's still time to fix them.
At EPAM I wrote connection tests early in the engagement that surfaced client-network issues before they could block deployment.
- 02
Quality is a release gate
Testing isn't a phase at the end. If the checks don't pass, it doesn't ship, however close the deadline is.
My WhatsApp agent ships only when all 138 eval runs pass in Arabic and English, and its booking rules are enforced in code, not left to the prompt.
- 03
Build for adoption, not the demo
A system is done when people rely on it. I design with the end users from requirements through rollout.
ESRA became the standard reporting procedure for all 800 NEOM contractors and cut compliance reporting from ~4 weeks to ~1 hour.
- 04
Make the next person faster
Automation, documentation and clear standards compound. I leave teams quicker than I found them.
At NEOM I cut developer onboarding from 6 weeks to 2 and application publishing cycle times by 75%.
05Writing
Latest · · 3 min read
A year of coding with agents: what actually changed
Twelve months ago I was skeptical. Here's how my day-to-day workflow shifted, and what I still refuse to hand over.
- Lessons from shipping AI agents to small businessesSeptember 2026
- You can't fix what you can't see: observability for agentsAugust 2026
- Voice agents are finally good enough. Are they good enough in Arabic?July 2026
06Contact
Two ways to work together. Pick the one that fits.
Hiring?
I'm open to senior AI engineering and solutions architecture roles where agents and data platforms need to work in production, not just in demos.
Need an AI agent for your business?
I build WhatsApp front-desk agents that answer customers in Arabic and English, book appointments, send reminders and hand off to your team when it matters.
