Marc Kahwaji

Marc Kahwaji

AI developer and full-stack engineer

I build multi-agent systems and the apps around them.

Software developer at HyperCycle, where I have put 15+ AI services into production. I work in Python, FastAPI, Docker, React, and TypeScript, from first prototype to running system.

See projectsDownload CVkahwajiimark@gmail.com

Open to full-time, contract, and freelance work. Based in Lebanon, relocating to France in September 2026.

PlanSearchRetrieveSummariseSynthesiseCritiqueReport

Agent layout from the research orchestrator project.

About

I am a software developer at HyperCycle, where I build AI microservices and multi-agent systems on decentralised infrastructure. On the side I take freelance work for clients who need AI features, a web app, or training for their team.

Most of my recent work is orchestration: systems of six or seven agents that hand work to each other, multi-tenant deployments that serve several organisations from one instance, and platforms with real-time AI support built in. I work across the whole stack, from Python backends and React or React Native frontends to Docker, CI/CD, and deployment.

I am in my final year of a BSc in Management Information Systems at the American University of Science and Technology, graduating in January 2027. My habit on every project is the same: get a working version out early, then improve it with real usage.

0

AI services in production

0

projects shipped

0

websites launched

Projects

Work from HyperCycle, freelance clients, and research. Client work is described without internal detail.

Multi-agent research orchestrator

Multi-agent system

  1. Plan
  2. Search
  3. Retrieve
  4. Summarise
  5. Synthesise
  6. Critique
  7. Report

A seven-agent pipeline that researches any topic and writes a cited report. A dashboard streams each agent's progress over SSE, with Prometheus metrics behind it.

Python, LangGraph, Groq, FastAPI, Next.js, Docker, SSE, Prometheus

Code

Crypto trading multi-agent system

Multi-agent system

  1. Market data
  2. News sentiment
  3. Signals
  4. Charts
  5. Summary

Five agents split the work of market analysis: price retrieval, news sentiment, trading signals with confidence scores, Plotly charts, and an executive summary. The web UI shows each agent as it runs.

Python, FastAPI, pyhypercycle_aim, Groq, Docker, Flask, Plotly, CoinGecko API, NewsAPI

Code

E-commerce customer support AIM

Multi-tenant AI service

Support agent with hybrid keyword and LLM routing across 26 FAQ categories, above 80 percent routing accuracy. Each tenant gets its own policies, responses, and sessions.

Python, FastAPI, pyhypercycle_aim, Docker

Quoz

Client platform, contract work

A B2B platform for managing bespoke art programmes for large commercial clients. I work on the application layer: approval workflows, design versioning, and automated PDF lookbooks.

React, TypeScript, Supabase, Cloudflare

AI-based asset classification for ISMS

Machine learning

Supervised model that classifies people and devices from log data (CERT r4.2), benchmarked against a rule-based baseline and mapped to ISO/IEC 27001 controls. Random Forest reached 97 percent accuracy against a 92 percent baseline.

Python, Random Forest, ANN, SVM

Dynamic AI-based Android malware detection

Systematic review

PRISMA-based review of 42 studies, screened from 1,724 papers, on runtime malware detection for Android: analysis techniques, models, datasets, adversarial robustness, and reproducibility.

PRISMA 2020, literature synthesis

Experience

  1. May 2025 to present

    Software developer, HyperCycle

    • Designed, built, and deployed 15+ production AI microservices (AIMs) with Python, FastAPI, and Docker on HyperCycle's decentralised infrastructure.
    • Built an internal AI tool that moved from pilot to daily production use and is now part of the core workflow of roughly 80 percent of company staff.
    • Architected multi-agent systems of six to seven coordinated agents (LangGraph, LangChain, CrewAI, Groq) and multi-tenant deployments serving several organisations with isolated data and custom configuration.
    • Implemented CI/CD pipelines (GitHub Actions and Aimifier) for automated deployment to Ubuntu nodes and hardened reliability across all services.

  2. April 2025 to present

    Freelance AI and software developer

    • Contract developer on Quoz, a B2B art programme management platform: approval workflows, design versioning, and automated PDF generation.
    • Built full-stack websites (React, Tailwind CSS, Node.js) with integrated AI backends and real-time chat.
    • Scoped and delivered CRM integrations, WhatsApp automation bots, AI-powered telephony systems, and machine learning solutions for dataset analysis and predictive modelling.
    • Designed and delivered technical training on LLMs, multi-agent systems, prompt engineering, and Python to learners from complete beginners to working developers.

  3. 2021 to January 2027, expected

    BSc in Management Information Systems, American University of Science and Technology (AUST)

    • Relevant coursework: AI and information systems, database management, software engineering, statistics.
    • Capstone: multi-agent research system, seven orchestrated agents with a real-time monitoring dashboard (LangGraph, Groq, FastAPI, Docker).

Skills

AI and machine learning
LangGraphLangChainCrewAImulti-agent orchestrationRAGprompt engineeringOpenAIGroqOllamaFAISSNLPcomputer visionpredictive modelling
Languages
PythonTypeScriptJavaScriptC++JavaBash
Backend
FastAPINode.jstRPCHonoBunREST APIsWebSocketswebhooks
Frontend and mobile
ReactNext.jsReact NativeExpoTailwind CSS
Data and infrastructure
PostgreSQLSupabaseSQLiteDockerGitHub ActionsAWS EC2LinuxNginxVercelPrometheus
Integrations
StripeWhatsApp Business API (Twilio)n8nZapierSMTP
Languages spoken
Arabic (native)English (fluent, IELTS C1)French (fluent)German (B1)

Contact

If you need an AI system built, a web app shipped, or a team trained, send a message. I reply within two working days.