Available for a 6-month end-of-studies internship · Jan 2027

Chehin Chaaben.

> Final-year ICT engineer at SUP'COM. I ship computer-vision models, data pipelines for messy real-world documents, and the Python backends that hold the whole thing together.

View selected work Get in touch
  • Python
  • PyTorch
  • pandas
  • FastAPI
  • SQL
  • Docker
  • C++
  • LangGraph
/selected_work

Projects that put ideas on a GPU.

A short tour of what I've been shipping: medical imaging, multimodal search, and a cognitive assistant that lives on your file system.

  1. 01 feb → may 2026

    Lung nodule detection on 3D chest CT

    An end-to-end pipeline for automatic detection of pulmonary nodules on volumetric scans, from lung segmentation to candidate scoring.

    • Lung segmentation on 3D CT with nnU-Net-derived models: Dice 0.72 ± 0.08.
    • Nodule detection evaluated at F1 0.58 to support malignancy triage.
    • PyTorch
    • nnU-Net
    • 3D CNN
    • CT imaging
    • Python
  2. 02 jan 2026

    Multimodal AI discovery platform

    A product search engine where you type a sentence or drop an image, and results come back in near-real time, with recommendations tuned to your history.

    • Text & image search using CLIP embeddings plus vector search for near-instant retrieval.
    • Personalized recommendations driven by preferences and search history.
    • CLIP
    • Qdrant
    • FastAPI
    • Python
  3. 03 mar 2026

    Cognitive assistant for your file system

    A document-research assistant that watches your file tree and keeps a vector index fresh in the background. Ask in plain language, get the file in seconds.

    • Real-time file-system indexing to a vector store, zero user action required.
    • Natural-language retrieval returns the right file in ~3 seconds.
    • RAG
    • Qdrant
    • Watchdog
    • LLM
    • Python
/services

What I can bring to your team.

Four areas where I've done the deep work and can contribute from day one.

Deep learning for computer vision

Model design and training on 2D & 3D data (segmentation, detection, embeddings), with honest evaluation on Dice, F1, and calibration.

  • PyTorch
  • nnU-Net
  • CNNs
  • 3D segmentation

Document AI & RAG pipelines

Multilingual ingestion (OCR, layout, tables), chunking that respects structure, hybrid retrieval and grounded answers.

  • Docling
  • OCR
  • BM25
  • Qdrant
  • GraphRAG

Local-first LLMs & agents

Multi-agent systems on small local models (SLMs), with a role per agent, hallucination guard-rails, and traceable event-sourced flows.

  • Ollama
  • LangGraph
  • MCP
  • Prompt engineering

Python backends & data plumbing

FastAPI services, SQLite/MongoDB stores, ETL for messy real-world documents, and the monitoring that makes any of it debuggable.

  • FastAPI
  • SQLite
  • ETL
  • Event sourcing
/about

Grounded, measured, shipped.

I'm a final-year ICT engineering student at SUP'COM, Tunis. Ranked 17th nationally out of prépa, now spending my days teaching GPUs to see and small language models to reason.

My taste in engineering is simple: build the smallest honest thing that works, then measure it. I like deep learning when the metric is real (a Dice score on real patient scans), and I like agents when they are forced to tell you why they said what they said.

Lately I split my time between medical imaging research, multimodal retrieval, and plumbing: the FastAPI + SQLite + Ollama stack that turns a notebook into something a team can actually poke at.

3+research projects shipped
3engineering internships
14document types in production pipeline
5specialized agents orchestrated
/experience

Where I've been plugging the wires in.

  1. 2026 sep → now

    EY Tunisia · Tutored project, Data & AI

    Tunis, Tunisia

    Intelligent document platform spanning text, tables, figures and images across multilingual sources.

    • Designed a 12-step ingestion architecture, from OCR to chunking into a SQLite catalog.
    • Built a pipeline covering 10 file formats and 14 document types in French, English and Arabic.
    • Shipped a POC benchmarking 5 extraction methods: Docling, LLMWhisperer, pdfplumber, raw & pre-processed OCR.
    • Python
    • Docling
    • EasyOCR
    • Tesseract
    • SQLite
    • BM25
    • RAG
  2. 2026 jun → aug

    Sofrecom Tunisia · Data & AI intern

    Sfax, Tunisia

    Partner-onboarding support platform. Turned five scattered manual lookups into one traceable, reversible event-sourced view.

    • Automated partner onboarding tracking: 5 manual searches collapsed into a single view, zero manual entry.
    • Designed a 5-agent architecture on local SLMs, each with its own role, context, anti-hallucination harness, and a portfolio-wide chatbot.
    • Integrated a remote inference engine under confidentiality constraints and built monitoring for data flow and agent performance.
    • Python
    • LLM/SLM
    • Prompt engineering
    • SQLite
    • ETL
    • Ollama
    • REST
    • Event sourcing
  3. 2025 jul → aug

    Arab Tunisian Bank (ATB) · Summer intern, IT Security

    Tunis, Tunisia

    Two-month summer immersion inside the IT security team of a Tunisian retail bank. My first look at how high-stakes production environments are actually defended day to day.

    • Observed the controls, audits and monitoring practices that keep core banking infrastructure trustworthy under regulation.
    • First professional exposure to a context where "it compiles" is nowhere near enough. Correctness, traceability and least privilege come first.
    • IT security
    • Banking
    • Monitoring
    • Compliance
  4. 2024 → 2027

    SUP'COM · U. of Carthage · Engineering degree, ICT

    Tunis, Tunisia

    Final-year student. 4-to-6 month end-of-studies internship starting January 2027.

  5. 2022 → 2024

    IPEIS · Preparatory cycle, Technology

    Sfax, Tunisia

    Classes préparatoires. Ranked 17th nationally.

/stack.yaml

Tools I reach for on autopilot.

Grouped the way I actually use them, not the way they look on a job post.

# ml & deep_learning

  • PyTorch
  • nnU-Net
  • CNNs
  • 3D segmentation
  • CLIP
  • Transformers
  • TensorFlow
  • scikit-learn
  • Model evaluation

# data & analytics

  • pandas
  • NumPy
  • Polars
  • Jupyter
  • Matplotlib
  • Seaborn
  • ETL pipelines
  • Event sourcing

# documents & rag

  • Knowledge graphs
  • GraphRAG
  • Docling
  • EasyOCR
  • Tesseract
  • BM25
  • Qdrant
  • Embeddings

# llms & agents

  • LangChain
  • LangGraph
  • Multi-agent
  • MCP
  • Prompt engineering
  • Local SLMs
  • Ollama

# backend & apis

  • Python
  • C++
  • SQL
  • FastAPI
  • REST APIs
  • SQLite
  • MongoDB
  • Pydantic

# dev & infra

  • Git
  • GitHub
  • Linux
  • Docker
  • Bash
  • CI/CD
  • VS Code
  • Postman
/contact.sh

Let's build something
that measures up.

I'm actively looking for a 6-month end-of-studies internship starting January 2027, with a focus on neural-network optimization for embedded systems. Always happy to talk about research, computer vision, or anything with a small local model inside.

chahin.chaaben@supcom.tn LinkedIn