Skip to content
ENFR

0x00 — index

Aymen Assaf

Data & AI Engineer

Building production machine learning systems, data pipelines, and low-latency backends. Dual background in Complex Systems Engineering (EILCO) and 1337 (42 Network).

8
production NER detectors
<50ms
RecSys feedback latency
0
measured RAG hallucination
−40%
manual ETL reduction

Aymen Assaf
node::aymen_assaf

0x01About

Completing an M2 in Complex Systems Engineering at EILCO alongside 1337 (42 Network). I focus on production AI systems, reliable data pipelines, and distributed backends. My work bridges applied machine learning (recommender systems, RAG/LLMs, computer vision) with solid software engineering (asynchronous services, C/C++, and Python). Looking for a 6-month Data & AI Engineer internship starting January 2027.

2026 — 2027

MSc (M2) in Complex Systems Engineering

EILCO — Université du Littoral Côte d'Opale — Calais, France

2024 — 2026

MSc in Data Science & Analytics

Université Cadi Ayyad — Morocco

Machine Learning, Deep Learning, time series, optimisation

2023 — present

Software Engineering — project-based curriculum

1337 UM6P (Réseau 42) — Morocco

Algorithms, systems programming, C/C++, network architectures

0x02Experience

Work experience

3 roles
[February — August 2026]

INNOVX

Casablanca, Morocco

Data Science / AI Intern

  • Designed and built an LLM-based platform for analysing and anonymising industrial contracts, deployed to a production environment (Docker, NGINX).
  • Built the full NLP pipeline: PDF text extraction, bilingual French/English named-entity recognition (spaCy, Microsoft Presidio) and 8 custom business-entity detectors.
  • Developed a multi-provider LLM orchestration layer (structured prompting, validated JSON outputs) and an asynchronous Python backend (FastAPI, Celery, Redis, PostgreSQL).

stack :: spaCy · Presidio · FastAPI · Celery · Redis · PostgreSQL · Docker · NGINX

[September 2023 — April 2024]

Grif Center GATechno

Safi, Morocco

Data Engineering Intern

  • Designed automated ETL pipelines (Node.js, SQL) cutting manual processing by about 40%.
  • Modelled data and built decision dashboards for activity monitoring.

stack :: Node.js · SQL · ETL

[May — August 2023]

SADELEC

Grenoble, France

Industrial Data Intern — IoT & time series

  • Processed industrial sensor data, structured logs in MongoDB and automated analysis reports in Python.

stack :: Python · MongoDB

0x031337 / 42

1337 (42 Network) — systems programming, concurrent network engines, and peer-reviewed software engineering.

1337 / 42Validated

ft_transcendence — real-time multiplayer platform & microservices

  • Server-authoritative WebSocket Pong game engine with real-time state synchronization, 1v1 matchmaking, and 4-player tournament brackets.
  • Dedicated AI opponent microservice (ai-service) featuring 3 difficulty tiers based on reaction latency and trajectory prediction.
  • Fastify microservices architecture with isolated SQLite stores (Prisma), NGINX reverse proxy, 42 OAuth login, and 2FA.
  • Full frontend rewrite in React 18, Vite, TypeScript, and Tailwind CSS with Socket.IO real-time messaging and presence management.

stack :: Fastify · React 18 · TypeScript · WebSockets · Socket.IO · Prisma · Docker · NGINX

Source codegithub.com/aymenssf/ft_transcendence
Real-time multiplayer match over WebSockets (60 fps physics)
Live GIF

Real-time multiplayer match over WebSockets (60 fps physics)

repo ↗

0x04Projects

Core projects & architectures

Problem

Recommend relevant movies by combining collaborative and content signals, while ingesting user feedback continuously, without batch retraining.

Architecture

SVD matrix factorization trained with multi-epoch online SGD, combined with content-based filtering. Distributed asynchronous orchestration with Ray, a Flask API surface, XMPP (JADE) inter-agent messaging, all containerized with Docker. Exploration is handled by an Epsilon-Greedy policy.

Usersratings & feedbackXMPP / JADEagentsRayonline SVD · contentε-greedyblendFlask APIserved recsfeedback loop <50 ms
PRISM CINE web interface — movie recommendations and online rating

feedback processing

<50 ms

stack :: Python · SVD · Epsilon-Greedy · Ray · Flask · XMPP / JADE · Docker

Source code

0x05Skills

34 skills · 42 connections

CL-01

RecSys & Search

Modeling, retrieval, and exploration — from real-time feedback ingestion (<50ms) to semantic disambiguation.

Skills
7
Connections
6

Recommender systems · SVD / matrix factorization · Content-based filtering · Hybrid RAG · SentenceBERT · Epsilon-Greedy

CL-02

NLP & LLM

Production NLP pipelines: extraction, Presidio anonymization, and structured LLM orchestration with validated outputs.

Skills
7
Connections
8

spaCy · Microsoft Presidio · GLiNER · Bilingual French/English NER · LLM Orchestration · Quantized Mistral-7B

CL-03

Systems & Backend

Concurrent architectures, real-time microservices, and containerized deployments behind NGINX.

Skills
10
Connections
8

FastAPI · Fastify · WebSockets · Docker · PostgreSQL / SQLite · Redis · Celery · Ray · C / C++ · Linux

CL-04

Data & Machine Learning

From raw data processing (−40% manual overhead) to image synthesis with generative models.

Skills
10
Connections
10

Python · PyTorch · scikit-learn · pandas · SQL · MongoDB · ETL Pipelines · Time series · CycleGAN / DDPM

Technical stack

Tools and frameworks applied across company internships, 1337 projects, and research.

TS-01

Languages

Python
SADELEC · PRISM CINE
SQL
Grif Center GATechno
C / C++
1337 / 42 curriculum
TypeScript
ft_transcendence
Node.js
Grif Center GATechno
TS-02

Machine Learning & AI

PyTorch
SatelliteGAN
scikit-learn
Data Science MSc curriculum
pandas
Data Science MSc curriculum
SVD / factorisation matricielle
PRISM CINE
CycleGAN
SatelliteGAN
DDPM
SatelliteGAN
Epsilon-Greedy
PRISM CINE
TS-03

NLP & LLM

spaCy
INNOVX
Microsoft Presidio
INNOVX
GLiNER
Hybrid RAG
SentenceBERT
Hybrid RAG
Mistral-7B
Hybrid RAG
RAG
Hybrid RAG
Wikidata API
Hybrid RAG
TS-04

Backend & Systems

FastAPI
INNOVX
Fastify
ft_transcendence
Flask
PRISM CINE
WebSockets
ft_transcendence
Socket.IO
ft_transcendence
Celery
INNOVX
XMPP / JADE
PRISM CINE
TS-05

Data & Storage

PostgreSQL
INNOVX
Redis
INNOVX
MongoDB
SADELEC
SQLite
Prisma
ft_transcendence
ETL
Grif Center GATechno
TS-06

Infra & Deployment

Docker
INNOVX · PRISM CINE · ft_transcendence
NGINX
INNOVX · ft_transcendence
Ray
PRISM CINE
GPU
SatelliteGAN
Linux
1337 / 42 curriculum
Git / GitHub
1337 / 42 curriculum
CI/CD
1337 / 42 curriculum

0x06Contact

Let's connect — open to Data & AI engineering roles and end-of-studies internships starting January 2027.

6-month end-of-studies internship · from January 2027 · On-site or Remote

assafaymen1@gmail.com

© 2026 Aymen Assaf — Data & AI Engineer

build f86adae · 2026-10-02 · 50.9481°N 1.8564°E

Next.js · Three.js · Tailwind CSS