Open to opportunities

> whoami

AI Engineer — agentic systems · RAG · LLM evaluation — shipping intelligence to production.

3+yrs building AI 24shipped works −90%QA effort automated 34+agents scored by GREAT
Explore the work Download résumé
[ 01 ]

Profile

// about
Portrait of Charbel Toumieh
ID: CT-2026 · AI ENGINEER Coventry, UK · UTC+0/+1

I build production-grade LLM systems — agents, RAG pipelines and evaluation frameworks that leave the notebook and survive real users.

Currently reading for an MSc in Applied Artificial Intelligence at the University of Warwick while leading AI engineering at Gabriel & Co. in New York — where I designed GREAT, an AI quality-assurance platform that replaced manual call review across 34+ agents and three channels.

My work covers the full arc of an AI product: discovery, prompt and agent architecture, MCP servers and tool integration, evaluation and A/B testing, and the dashboards that make it legible to the business. Before that — data products serving 100K+ users at CMA CGM, and Meta-funded multimodal research on paediatric ECG models.

  • practiceagents · RAG · evals · automation
  • nowAI & LLM Engineer @ Gabriel & Co.
  • studyingMSc Applied AI @ University of Warwick
  • timezoneUK-based · overlaps US & EU hours

[ 02 ] — flagship system · scroll to zoom

great — quality-assurance platform LIVE 
$ great report --scope=all
┌─ scope      34+ agents · 3 channels (inbound / outbound / email)
├─ replaces   manual call review
├─ qa effort  −90%
├─ eval quality +65% via structured outputs & A/B prompt testing
├─ rep scores 80s → 90s
└─ integrates 24+ internal tools · AWS data · dashboards
$ great stack
agents · MCP servers · RAG pipelines · evaluation frameworks · n8n · Node.js · PHP
$
Designed & led end-to-end @ Gabriel & Co., New York 2024 — present
[ 03 ]

Experience

// 2023 — present
01 Jun 2024 — Present New York, US · remote

AI & LLM Engineer

Gabriel & Co.

  • −90% QA effort. Led the development and evolution of GREAT — automated scoring across 34+ agents and three channels, lifting rep quality scores from the 80s into the 90s.
  • +65% evaluation quality. Designed evaluation frameworks, scoring methodologies and prompt architectures assessing customer interactions at scale.
  • 24+ tools integrated. Architected AI agents, MCP servers, RAG pipelines and workflow automation across internal tools, AWS data sources and dashboards.
  • Shipped AI for recommendations, retailer-brand monitoring, content generation and operational analytics.

llm-evaluation · prompt-architecture · mcp · rag · node.js · php · n8n

02 Jan 2026 — Jun 2026 Remote

Artificial Intelligence Researcher

Chehab Lab

  • 6+ encoders benchmarked. Researched Semantic Slice Selection for 3D medical imaging — BioMedCLIP, DINOv3, DenseNet, ResNet, MedVAE, RAD-DINO.
  • Designed experiments with cosine similarity, L2 distance, semantic-preservation and compression metrics to quantify representation collapse in medical AI models.

medical-imaging · vision-encoders · representation-learning

03 Jan 2025 — Oct 2025 Marseille, FR · remote

Data Strategy & Engineering

CMA CGM

  • +33% targeting accuracy. Enhanced clustering models for customer segmentation and route analysis.
  • 100K+ users served. Owned roadmap and delivery of two core data products across 100+ shipping routes and financial datasets.
  • −40% reporting time. Engineered ETL pipelines and automated dashboards covering 50+ KPIs.

data-products · clustering · etl · kpi-dashboards

04 Jan 2024 — Jan 2025 California, US · remote

AI & LLM Researcher

Meta-Funded Research Collaboration

  • Advanced a multimodal ECG model in a 7-researcher Meta-funded project — high-accuracy reconstruction and atrial-fibrillation prediction for paediatric care.
  • +20% reading accuracy. Improved AI-assisted ECG interpretation via evaluation, behavioural analysis, multimodal prompt experimentation and benchmarking.

multimodal-ai · healthcare · benchmarking

05 Nov 2023 — Feb 2024 Baabda, LB

Software Engineering Intern

BMB Group × PepsiCo

  • −30% package size. Built a .NET MAUI Blazor app with PepsiCo, adding offline Android SQLite support for stronger reliability and security.

.net-maui · blazor · sqlite · android

06 Jul 2023 — Oct 2023 Naqqache, LB

AI Engineering Intern

Intalio

  • 97% extraction accuracy. Engineered a Flask OCR API handling 100+ daily requests with YOLOv8, OpenCV and Tesseract — efficiency +35% via pipeline optimisation.

yolov8 · opencv · tesseract · flask

07 Mar 2023 — Jun 2023 Dbayeh, LB

Software Engineering Intern

eMcREY

  • −40% processing time. Created XML, Java and MySQL applications on Oracle platforms in a 4-person team, automating three core business processes.

java · mysql · oracle · xml

[ 04 ]

The Stack

// capabilities

hover / tap the core to decompose

ai / llm
AI AgentsMulti-Agent SystemsAgent OrchestrationRAGVector DatabasesSemantic SearchPrompt EngineeringStructured OutputsLLM EvaluationA/B Prompt TestingHallucination ReductionTool / Function CallingSelf-VerificationMCPLangChainLangGraphLlamaIndex
ml & data
Clustering & SegmentationModel EvaluationBenchmarkingExperimental ValidationSQL / SQLiteETL Pipelines
backend & apis
PythonFlaskNode.jsPHPJavaREST APIsAPI Integrationn8n
cloud & vision
AWSAzurePyTorchYOLOv8OpenCVTesseract OCREfficientNetGrad-CAMObject Detection
product
Product DiscoveryStakeholder ManagementKPI DefinitionExperimentationRoadmap Prioritisation
languages
English — full professionalFrench — full professionalArabic — native
[ 05 ]

Selected Works

// drag ⟷ scroll · 01 / 06
PROJ.01 — flagshipPython library

LoopGuard

AI-agent reliability framework

Framework-agnostic, zero-dependency loop detector for AI agents — semantic similarity, cycle detection and declarative guardrails for LangChain, LangGraph, CrewAI, Autogen and RAG pipelines.

python · semantic-similarity · guardrailsopen repo ↗
PROJ.02Explainable CV

DermSense

Skin-disease classifier with Grad-CAM

Medical vision pipeline contrasting an SVM/HOG baseline with a PyTorch EfficientNet-B0 transfer model — Grad-CAM heatmaps for diagnostic transparency, served via a responsive web UI.

pytorch · efficientnet-b0 · grad-camopen repo ↗
PROJ.03ML pipeline

OULAD Analytics

Student-retention forecasting, CRISP-DM

Retention analysis and success forecasting — XGBoost with RFECV feature selection, classifiers, regressors and unsupervised clustering (DBSCAN, GMM) behind a Python CLI.

xgboost · dbscan · gmmopen repo ↗
PROJ.04Systems

Commerce Microservices

Containerised FastAPI platform

E-commerce backend split into segregated customer and order services — HAProxy load-balancing, Prometheus scraping and Docker orchestration.

fastapi · docker · haproxy · prometheusopen repo ↗
PROJ.05Full-stack

Network Cell Analyzer

Android telemetry → live dashboard

Native Android diagnostics client capturing LTE/5G/Wi-Fi parameters, streaming telemetry through Flask APIs into a live server-side statistics dashboard.

android-sdk · flask · sqliteopen repo ↗
PROJ.06Algorithms

Supermarket Graph Analysis

Market-basket & path-finding engine

Logistics simulator modelling store layouts as weighted graphs — custom quicksort, DFS/BFS routing and association-rule analytics producing recommendations.

graph-theory · algorithms · testingopen repo ↗

Full archive 18 more

  1. 07Signal Processing StudiesMATLAB · DFT · BPSK/QPSKdata
  2. 08Public Health DashboardPython · Pandas · data vizdata
  3. 09CDC Clinical Classifierscikit-learn · clusteringai · ml
  4. 10Student Wellbeing InsightsFlask · SQL analyticsai · ml
  5. 11Cross-Platform MAUI AppC# · .NET MAUI · SQLiteapps
  6. 12Cipher WebFlask · classical cryptography · Azureweb
  7. 13Careology MarketplaceFlask · OpenAI chatbotweb
  8. 14Interactive Chart GeneratorMatplotlib · Seaborndata
  9. 15Jenkins Automation PipelineCI/CD · coverage assertionssystems
  10. 16TCP Proxy Server–Clientsockets · forwarding · cachingsystems
  11. 17Delay Multiplayer GameTCP sockets · real-time scoringsystems
  12. 18Weighted Graph ManipulationC++ · Dijkstra · MSTsystems
  13. 198-bit Arithmetic Logic UnitVHDL · digital logic · testbenchsystems
  14. 20Mini Library SystemFlask · cataloguing · search indicesapps
  15. 21Employee RegistryFlask · SQLite · schedulingweb
  16. 22Contact List ManagerDjango · CRUD · searchapps
  17. 23PyQt Desktop ClientPython · PyQt · desktop formsapps
  18. 24Flask User Registry APIFlask · CORS · RESTsystems
[ 06 ]

Education

// degrees & leadership
in progress — expected Sept 2026

MSc Applied Artificial Intelligence

University of Warwick

Coventry, United Kingdom

conferred

B.E. Computer & Communications Engineering

American University of Beirut

Beirut, Lebanon

> connect --with charbel

Let's build something
intelligent.

charbeltoumieh1@gmail.com
Write to me
Open to opportunities Coventry, UK · Beirut, LB US & EU time zones Roles · collaborations · research
GitHub ↗ LinkedIn ↗ Résumé ↓