Y.B.

I build AI agents that do real work — and dig into how they think.

I'm Yasseur Boutobba, an AI agents developer and final-year AI & Data Science student chasing the intersection of craft and intelligence — from agentic systems and RAG to full-stack engineering.

  • 01RoleAI Agents Developer
  • 02LocationAlgeria
  • 03FocusAI / Data Science, ESTIN
Download résuméOpen to opportunities

Articles

Notes on attention, interpretability, and where machine intelligence is heading.

Research interests

AI control & alignment
How skepticism toward untrusted information emerges in language models, framed within the AI control agenda.
Agents & RAG systems
Retrieval-augmented and multi-agent systems, end to end — with benchmarks to measure behavior.
Mechanisms & training
Transformers, training pipelines, and interventions like steering and instruction tuning.

The Stack

Every tool here is one I reach for by default.

AI & ML

  • NumPy
  • Pandas
  • Matplotlib
  • PyTorch
  • TensorFlow
  • Scikit-learn
  • AI Agents
  • LLM Integration
  • RAG Systems
  • LangChain
  • LangGraph
  • Vector DBs

Languages & Tools

  • Python
  • JavaScript / TS
  • FastAPI
  • Express.js
  • NestJS
  • Redis
  • Docker
  • SQL
  • Git
  • Linux
  • PostgreSQL
  • MongoDB
  • n8n
  • Google Colab
  • Hugging Face
  • Jupyter
  • Streamlit

Projects

Things built end to end — what each one is, what it runs on, and where the source lives.

  • Transformer From Scratch

    Implemented a decoder-only transformer in PyTorch for text generation and an encoder-decoder transformer for English–German translation, with a full training pipeline: custom attention, tokenization and mixed-precision training. Built end to end — custom SentencePiece tokenizers, memory-mapped datasets, teacher forcing, checkpointing, and training on both local (MPS) and cloud (Colab) hardware.

    • PyTorch
    • SentencePiece
    • Mixed precision
    • Google Colab
    • Python
    View source

    2026

    Personal project

  • Scholar — RAG Research Assistant

    Built a retrieval-augmented generation assistant that constructs a knowledge base from research papers to answer questions grounded in the source material. Implemented the retrieval pipeline — chunking, embeddings, vector search — and grounded generation to reduce hallucination when answering questions over the paper corpus.

    • RAG
    • Vector databases
    • Embeddings
    • LangChain
    • Python
    View source

    2026

    Personal project

Experience

Where I've worked, and what I studied — set down like two chapters in one record.

Work

  1. Jun 2026 — Present

    Agentic AI Developer

    DevFlowsRemote

    • Work within a team of engineers to design and deploy client-facing automation workflows in n8n and custom agentic systems in LangChain, covering the full path from requirements to production.
    • Connect AI workflows to client CRM/ERP tools and internal data sources, turning existing business processes into automated, multi-step pipelines.
  2. Apr 2026 — Jun 2026

    Data Science Intern

    SurvisionHybrid

    • Built a data cleaning pipeline to standardise and validate data, and produced business insights from it.
  3. Jun 2025 — Apr 2026

    Fullstack Developer

    RetailSpot / ProcessIQRemote

    • Built full-stack features across the frontend and backend using Next.js, Express.js, and MongoDB.

Education

  1. Sep 2022 — Present

    Engineer's Degree — AI & Data Science

    ESTINFinal year

    • Machine Learning, Data Science, Computer Vision
    • Probability & Statistics, Linear Algebra, Calculus
    • Time Series, Advanced Algorithms

Contact

Send a note and it lands straight in my inbox.