Available · AI Engineer @ Quicktext

Raed
Lazreg

AI Engineer · LLM & Agentic Systems

I ship

I build production multi-agent architectures and real-time voice agents that take live calls, check availability, and book reservations through tool calls, plus RAG pipelines that cut hallucinations in customer-facing products.

About

An engineer for systems that think, listen, and act.

I'm an AI Engineer focused on LLM and agentic systems, with production experience taking models from prototype to live, customer-facing deployments.

At Quicktext I architect multi-agent systems with LangGraph and the Model Context Protocol (MCP) for state management and context orchestration, and build a real-time hospitality voice agent on LiveKit that handles natural-language calls end to end.

Before that I engineered RAG pipelines and ran fine-tuning and prompt-engineering experiments that measurably improved accuracy and reduced hallucinations. I took 1st place at the AI Night Challenge with a computer-vision defect-detection system.

roleAI Engineer
focusAgents · Voice · RAG
stackLangGraph · MCP
based_inSousse, TN
languagesEN · FR · DE
studyingAI Eng. '26

Experience

Where I've shipped.

Two roles at Quicktext: first proving models in research, then putting agents into production.

AI Engineer · Quicktext

Jul 2025 - Present
  • Architecting and implementing advanced AI agent systems with LangGraph and the Model Context Protocol (MCP) for superior state management and context orchestration.
  • Experimenting with multi-agent architectures and agentic reasoning patterns.
  • Developing a high-performance hospitality real-time voice agent (LiveKit) that takes natural-language calls, checks live room availability, and books reservations through tool calls.

Data Scientist · Quicktext

Sep 2024 - Jun 2025
  • Engineered and deployed Retrieval-Augmented Generation (RAG) systems, enhancing LLM accuracy and mitigating hallucinations in customer-facing applications.
  • Executed fine-tuning strategies on large language models for domain-specific tasks and proprietary datasets.
  • Led experiments in advanced prompt engineering and chain-of-thought reasoning to improve model logical consistency.

Projects

Selected work.

From a national competition win to from-scratch model internals.

★ 1st placeComputer Vision

AI-Powered Ceramic Tile Defect Detection

Won 1st place at the 3rd edition of the AI Night Challenge (hosted by ARSII). A vision system that detects surface defects on ceramic tiles in real time.

OpenCVYOLOPython
Competition project · ARSII
ProductionVoice

Ultravox Voice Reservation Agent

A native speech-to-speech agent built on the Ultravox SLM for hospitality, with human-like voice interactions that manage bookings over a live audio stream.

Ultravox SLMWebSocketsFastAPIPython
Hospitality · real-time
NPMFrontend

Chatbot UI Widget

A reusable, published chatbot interface widget shipped as an NPM package: a drop-in conversational UI that integrates into external applications.

React.jsHTMLCSS
Published package
Model internalsSystems

microGPT

A minimal GPT implementation written from scratch in Rust, inspired by Karpathy's microgpt.py, building transformer internals at the metal.

RustTransformers
View on GitHub
MLMLOps

House-Price Prediction

House price prediction with a Random Forest model wrapped in a simple, reproducible MLOps workflow, from notebook to a clean training pipeline.

scikit-learnRandom ForestJupyter
View on GitHub
Open sourceMCP · Voice

Working in the open

Active with the toolchains behind my work, contributing to and tracking the MCP Python SDK / servers and the LiveKit agents framework.

MCP SDKLiveKit Agents
All repositories

Skills

The toolkit.

What I reach for, grouped by where it lives in the stack.

Agentic & LLM // orchestration

LangGraphModel Context ProtocolLangChainRAGMulti-agent systemsPrompt EngineeringFine-tuningChain-of-Thought

Real-time & Voice // interfaces

LiveKitUltravox SLMWebSocketsFastAPITool calling

Languages // code

PythonRustTypeScriptJavaScriptSQL

ML, CV & Data // foundations

OpenCVYOLOscikit-learnRandom ForestsReact.jsGitOracle Cloud (OCI)

Education & credentials

Foundations.

2023 - 2026

AI Engineering Degree

École Polytechnique de Sousse Data science · Computer vision · Python
certification

Data Scientist with Python

DataCamp Career track
certification

OCI AI Foundations Associate

Oracle Cloud Infrastructure Cloud AI services
languages & community

EN · FR (DELF B2) · DE (ÖSD B1)

Former Treasurer, Espita Google Club Trilingual · leadership

Contact

Let's build something that thinks.

Open to AI engineering roles and collaborations on agentic systems, voice AI, and RAG. The fastest way to reach me is below.

// AI Engineer focused on agents, voice, and RAG.
Currently building at Quicktext.