Curriculum Vitae

Ralph Touma

Machine Learning Engineer & Data Scientist

Based in
Paris, France
Email
ralphtouma@outlook.com
LinkedIn
in/toumaralph
Portfolio
ralphtouma.com

Summary

Machine learning engineer and data scientist specialising in federated learning, forecasting and retrieval-augmented generation. Pairs research, including a co-authored 2026 paper on clustered federated learning, with production delivery for more than 15 clients. Holds three AWS certifications and works in French, English and Arabic.

  • 15+Clients delivered
  • R² 0.97Mesh-routing model
  • 3AWS certifications
  • 1Co-authored paper

Experience

Web Developer

BYND Network · Jdeideh, Lebanon

  • Delivered full-stack websites for 15+ agency clients, owning front end, back end, database and deployment.
  • Increased a client's conversion rate by 30% through analysis of visitor behaviour.
  • Built and customised Shopify storefronts using Liquid, JavaScript and the Shopify API.
  • Developed custom Shopify apps in Node.js to extend store functionality.
  • Engineered bespoke React and Vue web applications for clients outside the Shopify ecosystem.

Case study: ralphtouma.com/bynd-network

Freelance Full-Stack & AI Engineer

Independent · Paris, France and remote

  • Deliver full-stack web applications end to end, from scoping and architecture through front end, back end, database, deployment and handover.
  • Build LLM-powered assistants, retrieval-augmented generation over private documents, tool-using agents and workflow automation integrated into clients' existing products.
  • Own every client relationship directly, translating business goals into scoped technical plans, production code and measurable results.
  • Deploy models and agents on AWS and Azure, working across React, Vue.js, Node.js, Express, Python, SQL and MongoDB.

Machine Learning Research Intern

Green-Communications · Paris, France

  • Developed ML policies and a forward model (R² ≈ 0.97) to configure wireless mesh network routing, replacing an approach that could not work.
  • Validated configurations on 10-, 28- and 104-node networks, and resolved the 104-node shortfall with domain-informed constraints.
  • Implemented a three-layer clustered federated learning framework for soil-moisture sensors, simulated on 80 sensors.
  • Achieved RMSE 2.43 after 12 readings in cold-start tests, against 3.18 for local models and about 3.2 for a single global model.
  • Co-authored a research paper on the framework, submitted in 2026.

Case study: ralphtouma.com/green-communications

Backend Developer Intern

Eurisko · Adma wa Dafneh, Lebanon

  • Engineered back-end services in Node.js and Express.js, designing the server-side logic and endpoints behind the front end.
  • Optimised application performance and responsiveness across the request cycle.
  • Modelled and managed application data in MongoDB.
  • Collaborated with front-end developers in a Git-based team workflow.

Search Engine Evaluator & Data Collector

Appen · Freelance, remote

  • Evaluated the accuracy and relevance of search results returned by virtual assistants including Siri, Alexa and Google Assistant.
  • Extracted and classified business information from media sources to structure data for AI training datasets.

Education

Master's in Computer Science

Université Paris-Saclay · Paris, France

Master's in Data Science

Saint Joseph University of Beirut · Lebanon

Bachelor's in Computer Science

Notre Dame University · Lebanon

Research & Projects

Publication

A Three-Layer Game-Theoretic Framework for Soil-Moisture Clustered Federated Learning. Co-author. Research conducted at Green-Communications.

Enterprise RAG Assistant for Technical Documents

Retrieval-augmented generation

  • Built an ingestion pipeline that chunks, embeds and stores PDFs and internal documents as vectors for semantic retrieval.
  • Combined semantic search with lexical matching and relevance ranking, so exact terms such as part numbers and function names still retrieve the right passage.
  • Grounded every answer in citations to the source documents, making each answer verifiable.

Case study: ralphtouma.com/enterprise-rag-assistant

Privacy-Aware Federated Weather Forecasting

Time-series forecasting

  • Built next-day forecasting across city-level clients on Flower's simulation framework; each city trains on its own station records and shares model updates, never raw observations.
  • Personalised the model for each city, so every client keeps a model adapted to its own conditions.
  • Benchmarked centralised, local-only, FedAvg and personalised federated models on the same data.
  • Tracked per-client and global error across rounds with custom metric aggregation.

Case study: ralphtouma.com/federated-weather-forecasting

SUPER FOLIO

Playable résumé

  • Built an original 8-bit side-scrolling platformer from scratch in vanilla JavaScript and HTML5 canvas, with no game engine, libraries or build step.
  • Implemented a fixed 60 Hz simulation, with pixel art painted from character grids at load and music synthesised live in the browser.

Play it: play.ralphtouma.com

Skills

Core competencies
End-to-end ML pipelinesforecasting modelsdata preprocessingexperiments and validationanalytics dashboards
Programming
PythonRSQLJavaScriptTypeScriptJavaC++C#
Data & ML
pandasNumPyscikit-learnTensorFlowPyTorchHugging FaceAmazon SageMakerJupyter
Cloud
AWSMicrosoft Azure
Visualization & BI
Power BILookerExcelMatplotlibSeaborn
Databases
MySQLMongoDB
Web & APIs
ReactVue.jsNode.jsExpress.jsREST APIsHTMLCSS

Certifications & Languages

AWS Certified Generative AI Developer – Professional

Amazon Web Services · Early adopter, among the first 5,000 to earn it · Verify

AWS Certified Machine Learning Engineer – Associate

Amazon Web Services · Verify

AWS Certified AI Practitioner

Amazon Web Services · Verify

Languages

French, English, Arabic