Skip to content

Platform & AI Engineer — Rabat, Morocco

From infrastructureto interface.

I build production systems across the whole stack: conversational AI and multi-model orchestration, multi-tenant Java backends, CRM and messaging integrations, custom protocol servers, and the client-facing web on top of them. Ten enterprise brands, eight channels, running in production.

Open to remote (EU / US-overlap) and relocationSee the work

2×

lead-to-visit conversion against a human baseline

$0.10–0.80

per qualified lead, against paid acquisition

10+

enterprise brands on one platform

8

channels served from one platform

01 — Selected work

Five production systems.

02 — How it works

What happens when a customer messages.

One platform, ten enterprise brands, eight channels. This is the path a single message takes, and the part of it I built.

  1. Illustration: a phone showing a chat thread, its messages routed through three paths that converge into a single highlighted block.

    Step 01

    The message arrives

    A customer writes on WhatsApp, Instagram or Messenger. All eight channels land in one conversation, so nobody is checking three different inboxes.

    • Meta WhatsApp Cloud API
    • Official coexistence — the Business app and the API share one number
    • Template and opt-in compliance
    • 100+ webhooks per second in production
  2. Illustration: a source node feeding several parallel rails; one rail breaks and the flow steps down to the next, still reaching the end.

    Step 02

    The agent answers

    An AI agent qualifies the lead and replies in the customer’s own language, at any hour, and hands the qualified ones to the sales team with the context attached.

    • Agentic function calling against the CRM and catalogue
    • Hybrid retrieval — BM25 and HyDE over pgvector
    • Handles Darija and French mixed mid-sentence
    • A nightly LLM judge scores conversations and gates releases
  3. Illustration: a stack of chat bubbles feeding a structured record whose fields are being populated.

    Step 03

    It becomes a record

    The conversation lands in the client’s CRM as a structured lead or order with the fields already filled in. The sales team keeps working where they already work.

    • Zoho CRM and Salesforce pipelines over OAuth2 and HMAC
    • Per-tenant field mapping
    • Idempotent webhooks — a redelivery never duplicates a record
    • Click-to-WhatsApp ad attribution carried into the record

58%vs 30%

lead-to-visit conversion, measured against a human team working office hours on the same funnel

03 — Live

Shipped and live.

safq.ai

SaaS product serving six enterprise clients. Stripe billing, NextAuth v5, Prisma, and a dashboard over the AI orchestration backend. Built and run solo.

  • SaaS
  • Stripe
  • Product
Visit site ↗

useinvestment.com

Growth diagnostics for Shopify brands. Reads a store’s public catalogue and returns findings with the evidence attached, benchmarked against named competitors.

  • Product
  • Diagnostics
  • Next.js
Visit site ↗

04 — Capabilities

Areas of expertise.

01

AI systems

Deploying large language models in production environments with the availability, accuracy and validation guarantees enterprise clients require.

  • Model orchestration on Gemini
  • Agentic function calling
  • Hybrid RAG — pgvector, BM25, HyDE
  • Custom MCP servers
  • Output validation & guardrails
  • TTS, transcription, vision
  • Darija & multilingual support

02

Backend & data

Multi-tenant backend architecture with strict data isolation, comprehensive automated test coverage and measured production reliability.

  • Java 17 / 21 — Oracle Certified Professional
  • Spring Boot 3, Spring Security
  • Virtual threads
  • JPA / Hibernate performance
  • PostgreSQL, pgvector, Redis
  • Flyway, Quartz, Resilience4j
  • Apache Kafka, event-driven pipelines
  • REST / OpenAPI design

03

Integrations & CRM

Connecting messaging platforms, CRMs and payment providers into unified pipelines, with the resulting data available where sales teams already work.

  • Meta WhatsApp Cloud API
  • Official WhatsApp coexistence
  • Click-to-WhatsApp ad attribution
  • Zoho CRM extensions, widgets & functions
  • Salesforce
  • Template & flow automation
  • Webhooks, OAuth2, HMAC
  • Stripe

04

Product & delivery

End-to-end delivery across interface, infrastructure, domain configuration and deployment, including direct technical discovery with clients.

  • Next.js, React, TypeScript
  • Tailwind, motion design
  • Angular, Vue
  • Docker, Railway, Vercel, Nginx
  • CI/CD, Jenkins
  • DNS & deployment architecture
  • JUnit, Mockito, Playwright
  • Technical discovery with clients

05 — Experience

Professional experience.

  1. Voom Digital

    Now

    Platform & AI Engineer

    Jan 2026 — Present · Casablanca, hybrid

    Conversational AI, platform engineering and CRM integrations for a multi-tenant messaging platform serving 10+ enterprise tenants across eight channels.

    • Built a production AI agent with seven agentic tools, serving 10+ tenants and converting 58% of leads to visits against 30% for a human team on office hours.
    • Implemented official Meta coexistence, allowing clients to retain the WhatsApp Business app on their own devices while the same number operates through the Cloud API, with message history synchronised in both directions. Previously, connecting a number to the API meant deleting the client’s WhatsApp accounts outright.
    • Extended the agent to Instagram and Messenger, and added real-time catalogue and order sync so it can assemble and submit e-commerce orders in-chat on customer confirmation.
    • Delivered custom Zoho CRM widgets surfacing a contact’s live conversation within the CRM record, eliminating tab switching and manual lookups for sales agents.
    • Captured Click-to-WhatsApp advertisement webhooks and injected the originating ad context into the agent prompt, so initial responses reflect the specific campaign the customer engaged with.
    • Automated per-client workflows dispatching approved WhatsApp templates and flows from CRM triggers, with variable resolution and delivery tracking.
    • Integrated the AI agent with the CRM prospect module for automated prospect creation from qualified conversations.
    • Ran the backbone on Java 21 virtual threads and Spring Boot 3.3, sustaining 100+ webhooks/sec.
    • Java 21
    • Spring Boot 3.3
    • PostgreSQL / pgvector
    • Redis
    • Gemini
    • Claude
    • OpenAI
    • MCP
    • Meta Cloud API
    • Zoho CRM
    • Salesforce
    • Next.js
    • Docker
  2. Orange Morocco

    Software Engineer Intern — Final Year Project

    Mar — Sep 2025 · Rabat

    Digitalised the FabLab end to end, replacing manual paperwork for 50+ members and 200+ users.

    • Modular microservices backend of 9+ Spring Boot services with inter-service messaging over Apache Kafka.
    • Real-time reservation system using optimistic locking, eliminating 100% of scheduling conflicts.
    • Integrated 10+ IoT devices — 3D printers, Raspberry Pi — over REST, plus an AI support assistant.
    • Java 17
    • Spring Boot 3.4
    • Angular
    • MySQL
    • Apache Kafka
    • Swagger
  3. Lanoria Club

    Backend Java Developer, Intern

    2024 · Rabat, hybrid

    Backend for an HR platform serving 50+ users, built and documented end to end.

    • Built 25+ REST endpoints secured with Keycloak SSO and role-based access control.
    • Optimised queries and the cache layer, cutting latency by 40%.
    • Full Swagger/OpenAPI documentation, and delivery presentations to the client.
    • Java 17
    • Spring Boot 3.1
    • PostgreSQL
    • Keycloak
    • Angular
    • Swagger
  4. YOS Études

    Full-Stack Developer, Intern

    2023 · Rabat, hybrid

    Secure HR portal with granular permissions and live performance dashboards.

    • Three-level RBAC with granular permission management across the portal.
    • Real-time dashboards tracking 12+ KPIs over WebSocket.
    • Java 17
    • Spring Boot 3.1
    • Spring Security
    • JWT
    • Angular
    • SQL Server
  5. Ministry of National Education

    Full-Stack Developer, Intern

    2022 · Rabat

    School management platform for the Moroccan national education ministry.

    • Designed and built a platform for administering schools, from data model to interface.
    • Java
    • Spring
    • JWT
    • Design patterns
    • UML

06 — Background

Engineer, Rabat.

Software engineer specialising in production AI systems, backend architecture and enterprise integrations. Oracle Certified Professional in Java SE 17, holding a State Engineering degree in Computer Science from EMSI Rabat.

At Voom Digital I overhauled a multi-tenant conversational platform that now runs across eight channels for more than ten enterprise tenants.

I also designed and built safq.ai from scratch, the product that runs over that platform: its architecture, data model, deployment and infrastructure. It serves six enterprise clients today.

Day to day that means multi-model LLM orchestration, hybrid retrieval, Model Context Protocol servers, and integrations across Zoho CRM, Salesforce and the Meta WhatsApp Cloud API. I also design and build the interfaces clients actually look at.

Education

  • State Engineering Degree in Computer Science

    EMSI Rabat

    2022 — 2025

  • Higher Technician Diploma (HND), Software Development

    ISTA NTIC

    2019 — 2021

Certifications

  • Oracle Certified Professional: Java SE 171Z0-829
  • Oracle Database SQL Certified Associate1Z0-082
  • IBM Full-Stack Developer
  • Google Cloud Java Microservices
  • AWS DevOps

Languages

  • ArabicNative · C2
  • EnglishFluent · C2
  • FrenchProfessional · C1

07 — Contact

Available for new roles and projects.

Open to remote positions with EU or US overlap, and to relocation. For hiring enquiries, please include the scope of the system and the timeline.