FIELD NOTES 202601 — PROFILE

RESEARCH & ENGINEERING

Ali Ncibi

AI Engineer · PhD in Computer Science / AI

I build machine-learning systems that move from careful research into reliable software. My work spans activity recognition, NLP, LLM infrastructure, and agentic applications.

Portrait of Ali Ncibi
FIG. 01Ali Ncibi
Research
Human activity recognition · NLP · machine learning
Engineering
LLM systems · data pipelines · infrastructure
Based
Massy, Île-de-France · available for opportunities

From research questions to working systems.

I’m an AI engineer with a PhD in computer science. My work brings together machine learning, natural-language processing, and practical software engineering—from studying how sensor data represents human activity to building and evaluating LLM applications.

I care about sound experiments, transparent evaluation, and systems that remain useful beyond a prototype. I’m currently based in Île-de-France and available for new opportunities.

Machine learning, grounded in evidence.

Dec 2022 – Jun 2026

Saclay, France

PhD Research Contract · Inria

Led R&D on AI-driven human activity recognition in smart environments, from heterogeneous sensor data to reproducible model evaluation.

  • Structured 13 public sensor datasets and built a containerised machine-learning pipeline.
  • Evaluated classical, deep-learning, and text-based feature approaches with rigorous cross-validation.
  • Benchmarked 7 event-transcription strategies and 3 embedding models; observed up to 30% better cross-domain recognition in specific settings.

Nov 2021 – May 2022

Paris, France

Data / NLP Engineer · SouthPigalle

Developed NLP components for a jewellery-shopping chatbot with data science and cloud teams.

  • Created a template-based slot-filling corpus and an interactive data-visualisation interface.
  • Fine-tuned a BiLSTM-CRF tagger, exported it to ONNX, and containerised its inference API for production integration.
  • Improved average balanced accuracy by 7% per tag.

Mar 2021 – Sep 2021

Clamart, France

NLP / Machine Learning Intern · INSERM

Worked with medical teams on clinical timeline reconstruction from medical reports.

  • Defined a temporal-relation annotation scheme and annotated reports.
  • Built text-tagging and classification models for medical concepts and temporal relations.
  • Developed an interactive patient-timeline visualisation tool.

Building across the AI stack.

Wassla

AI-powered multilingual news platform

Aug 2026 – Present

A trilingual English, French, and Arabic news pipeline. LangGraph agents translate, categorize, summarize, and independently review articles before they reach an SSR reader with semantic search and citation-grounded answers.

  • LangGraph
  • LLMs
  • RAG
  • pgvector
  • Nuxt
Source code

Self-hosted LLM platform

Autoscaling inference on Kubernetes

Sep 2026 – Present

A reproducible, self-hosted LLM stack on OVH Managed Kubernetes with vLLM, Open WebUI, PostgreSQL/pgvector, Redis, and Prometheus/Grafana. Load tests validated scaling from one to three inference replicas.

  • Kubernetes
  • vLLM
  • PostgreSQL
  • Redis
  • Prometheus
Source code

Research in human activity recognition.

  1. Towards Domain-Robust Activity Recognition using Textual Representations of Binary Sensor Events

    18th International Conference on Agents and Artificial Intelligence (ICAART)

    Marbella, Spain · March 2026 HAL ↗

  2. A Framework for Data-Driven Development of Human Activity Recognition Systems using Binary Sensors

    21st International Conference on Distributed Computing in Smart Systems and the Internet of Things

    Lucca, Italy · June 2025 HAL ↗

  3. Une plateforme uniformisée pour la reconnaissance des activités humaines à partir de capteurs binaires

    25e Conférence francophone Extraction et Gestion des Connaissances (EGC)

    Strasbourg, France · January 2025 HAL ↗

A foundation in engineering and AI.

PhD in Computer Science / AI

Université de Versailles–Saint-Quentin · Paris-Saclay

Research in machine learning, human activity recognition, and AI-driven data processing. Successfully defended 1 June 2026.

2022 – 2026

Engineer in Data Science / AI

Télécom Paris · Institut Polytechnique de Paris

2019 – 2021

Engineer in IT & Telecommunications

École Nationale d’Ingénieurs de Tunis

2017 – 2019

Tools I work with.

AI & machine learning

Machine learning · Deep learning · NLP · LLMs · Agentic systems · RAG · Evaluation

Frameworks & languages

Python · SQL · PyTorch · TensorFlow · scikit-learn · Hugging Face · LangGraph

Data & services

PostgreSQL · pgvector · FastAPI · Pandas · ONNX · OpenAI-compatible APIs

Infrastructure

Docker · Kubernetes · vLLM · GitHub Actions · Prometheus · Grafana · OpenTelemetry

Languages

French — bilingual · English — C1 · Arabic — native

Thinking out loud.

March 2026 · 4 min read

From binary sensor events to textual representations

A short introduction to my PhD research on using language-model representations to improve human activity recognition across environments.

All notes

Have a research or engineering problem in mind?

contact@alincibi.page