Ismail Ben Alla

Ismail Ben Alla

Data & AI Engineer at Dailymotion · Paris, France

I build deep learning systems from scratch. Agentic AI · RAG · LLM evaluation. Then I ship them as real products.

I’m an AI engineer at Dailymotion in Paris, where I build data and AI pipelines on GCP: LLM-powered enrichment, multilingual retrieval, evaluation against human-reviewed golden datasets, and MCP-based agents. Before that I built document-extraction agents at Monce.ai, real-time voice conversion at SteelSeries, and multilingual legal RAG at Juridia.

I studied artificial intelligence at École Centrale Lyon (MSc) and computer science at ENSA Fès, where I graduated top of my class. In my own time I rebuild deep learning systems from scratch — GPTs, Vision Transformers, autograd — because I don’t understand something until I’ve built it.

Portrait of Ismail Ben Alla

What I’m building

Products I develop on my own time and want to put in people’s hands.

Experience

  1. Data & AI Engineer

    Jun 2026 – Present

    Dailymotion · Ray AI Platform · Paris, France

    • Built Airflow pipelines using Gemini with Google Search grounding to enrich web, app, and CTV sitelists with descriptions, IAB categories, and language metadata.
    • Developed BigQuery-to-Qdrant ingestion and multilingual retrieval with English-normalized context, semantic search, and LLM reranking.
    • Designed web-scraping workflows and LangSmith evaluation using human-reviewed golden datasets and ranking metrics.
    • Developed an LLM-powered Alert Agent that pairs Gemini with MCP clients and dedicated Airflow, GCP, and BigQuery MCP servers to investigate pipeline, Dataflow, Pub/Sub, and SQL incidents from Slack alerts.
    • Led a data-lineage technology assessment across BigQuery, Airflow, dbt, Pub/Sub, Dataflow, GCS, and BI, and prototyped an OpenLineage runtime integration.
  2. AI Engineer

    Sep 2025 – Jun 2026

    Monce.ai · LLM-powered data extraction · France

    • Designed a human-in-the-loop feedback system for continuous agent improvement, increasing document-extraction accuracy by about 18%.
    • Built a modular agent evaluation framework covering 10+ metrics, including precision, recall, hallucination rate, and latency.
    • Refactored LangChain / LlamaIndex extraction workflows with asynchronous processing, reducing average processing time by about 35%.
  3. AI Engineer

    Apr 2025 – Sep 2025

    SteelSeries · Real-time voice conversion · France

    • Built a PyTorch and ONNX voice-conversion system with end-to-end latency below 80 ms.
    • Developed real-time multi-voice blending on top of the LLVC architecture.
    • Improved MOS by 0.3 points over the StreamVC baseline through targeted fine-tuning and waveform post-processing.
  4. AI Engineer

    Jan 2024 – Jul 2024

    Juridia · RAG-powered legal AI · Morocco

    • Designed and deployed an Arabic / French RAG system with FastAPI, LlamaIndex, PostgreSQL, and Qdrant, with average query latency below 350 ms.
    • Improved retrieval precision by 18–22% with cross-encoder reranking, BM25 + dense hybrid search, and multilingual embeddings.
    • Built a search interface supporting keyword, semantic, and metadata search across 100k+ legal documents.
  5. Independent AI Engineer

    2024 – Present

    Qlaver · AI email-management agent · Remote

    Details
    • An agent that summarizes email to each user’s preferences and delivers the summary as audio.
    • Private project: I see real potential in it, so the code and the details stay closed for now.
  6. Computer Vision Engineer (Intern)

    Aug 2023 – Sep 2023

    3D Smart Factory · Autonomous-driving perception · Morocco

    Details
    • Developed YOLO and SegNet models for object detection and semantic segmentation in autonomous-vehicle images.
    • Trained and optimized the models in PyTorch on Kaggle GPUs with the Mapillary Vistas dataset, then deployed them on FPGA with Vivado.
  7. Machine Learning Engineer (Intern)

    Jul 2023 – Aug 2023

    3D Smart Factory · LLM pre-training & evaluation · Morocco

    Details
    • Pre-trained and fine-tuned GPT-2 on a custom scientific corpus for academic Q&A generation.
    • Built reproducible training with PyTorch Lightning and Weights & Biases, evaluated with BLEU, ROUGE, and perplexity.
    • Generated training data with the OpenAI API and cleaned text with NLP tools such as Hunspell, SymSpell, FuzzyWuzzy, TextBlob, and spaCy.
  8. Data Science Intern

    Jun 2023 – Jul 2023

    Technocolabs Softwares · Credit-risk analytics · India (remote)

    Details
    • Built a predictive analysis pipeline for early-repayment risk on bank loans with TensorFlow, Pandas, and NumPy.
    • Applied ANOVA, regression analysis, chi-squared and hypothesis tests, PCA, RFE, and mutual information for feature analysis and selection.
    • Handled preprocessing, exploratory analysis, and visualization with Matplotlib and Seaborn.
  9. Web & AI Developer (Intern)

    Aug 2022 – Oct 2023

    Dev-Bay · Web platform & recommendations · Morocco

    Details
    • Developed the official Dev-Bay website, improving its features and responsive design.
    • Built a recommendation system based on user behaviour using a classification model.

Selected projects

See all 37 projects →

Lab notes

All lab notes →

Education

  1. MSc in Artificial Intelligence (Research)

    Sep 2024 – Oct 2025

    École Centrale Lyon · Lyon, France

  2. Computer Science Engineering Degree — top of the 2024 class

    Sep 2019 – Aug 2024

    National School of Applied Sciences (ENSA) Fès · Fès, Morocco

Leadership & community

  1. Project Leader

    2021 – 2024

    IEEE ENSA Fès Student Branch

    Coordinated student projects in AI and robotics, managed teams, and organized scientific events.

  2. Co-founder

    2022 – 2024

    INFO-CHAIN — Computer Science & Blockchain Club

    Organized and led workshops on blockchain, collaborative development, and student leadership.

  3. Web Project Leader

    Sep 2021 – Feb 2023

    JEEF — Junior Enterprise of ENSA Fès

    Supervised web projects for clients, managed partner relations, and coordinated project communication.

Skills

Agentic AI

Tool calling · MCP clients & servers · Multi-agent workflows · Structured outputs · Guardrails · Human-in-the-loop · LlamaIndex

LangChainLangGraph

RAG & Retrieval

Hybrid search (BM25 + dense) · Multilingual embeddings · Cross-encoder reranking · LLM reranking · Grounding · Qdrant · PGVector

LLM Evaluation & LLMOps

Golden datasets · Offline / online evaluation · Precision / recall · Hallucination metrics · LangSmith

MLflowWeights & Biases

Deep Learning

Transformers · Vision Transformers · Fine-tuning (LoRA) · Multimodal models · Real-time audio (StreamVC, LLVC) · Graph neural networks

PyTorchPyTorch LightningONNX

Data & Cloud

Azure AI Foundry

GCPBigQueryAirflowDataflowPub/SubAWSPostgreSQLDockerKubernetesTerraformGitHub Actions

Languages & Web

SQL

PythonTypeScriptFastAPIReactNext.js

See my tech stack in 3D, with what I built with each tool →

Awards & languages

Contact

The best way to reach me is by email. I’m happy to talk about AI engineering, my projects, or research ideas.

All the ways to reach me →