Alexandre Hennequin

Curriculum vitae

Alexandre Hennequin

AI Consultant · Data Scientist · Applied Researcher — Marseille, France — working remotely

Summary

Independent AI & Data Science consultant. I help organizations ship LLM applications, retrieval systems, and data pipelines that are reliable, measurable, and cost-conscious. Comfortable from empty repo to production deployment: data engineering, embedding pipelines, vector search, LLM application architecture, and the observability that keeps them honest.

Experience

Independent AI & Data Science Consultant

Self-employed

Clients across real estate, land administration, and creative media need production-grade LLM systems in place of slow manual document work — with answers that are traceable, reliable, and fast.

  • Designed and deployed retrieval-augmented generation (RAG) agents on Qdrant vector stores, with citation-backed answers
  • Built end-to-end LLM products: a real-estate RAG assistant and chatbot, a RAG chatbot that drafts Word documents, and a children's story generator (LLM + TTS) delivered as a mobile app
  • Owned projects end-to-end: scoping, architecture, implementation, deployment, and client handover

Impact

  • Cut multi-document retrieval from 15–20 minutes to under a minute for a land-registry client
  • Shipped 3 production LLM products across real estate, document drafting, and mobile storytelling

Lead Data Scientist

O-Kidia

O-Kidia set out to build new technology to assess mental health in children through a holistic view — captured via handheld tablet sensors displaying "games" that were actually cognitive tests. As its first employee, I built the data and ML foundation the product ran on.

  • As first employee, laid the foundation for data handling and processing
  • Built, trained, and fine-tuned models handling multiple modalities of data
  • Worked across multiple disciplines: clinic, ML/AI, and IT security for clinical data and GDPR
  • Prepared for ISO 13485 certification as a medical device

Impact

  • Processed ~1 TB of multimodal assessment data in near-real time — roughly 10x faster than manual annotation
  • Contributed to 2 peer-reviewed articles and multiple conference presentations

Data Scientist

CrocosGoDigital

Built a video processing pipeline for facial emotion recognition, from raw footage to interpretable results, running on a single cloud GPU.

  • Built a facial emotion recognition pipeline in PyTorch running on a single cloud GPU
  • Displayed results in a visual dashboard presented at a conference

Impact

  • Cut video processing time 5x for a facial emotion recognition pipeline on a single cloud GPU

Skills

AI / LLM

RAG systemsPrompt engineeringLLM agentsEvaluationRetrieval optimization

Data / ML

PythonPandas / PolarsScikit-learnPytorch / TensorFlowEvaluation & metricsExperimentation (MLFlow)

Data Engineering

AirflowSQLETL/ELTQuality monitoring

Research & Methodology

Experimental designRigorous evaluation methodologyStatistical analysisPeer-reviewed researchPhD in cognitive science

Infrastructure / Ops

DockerVector stores (Qdrant)Cloud (AWS)CI/CDObservability (LangSmith)

Education

Google Scholar

Ph.D. in Cognitive Science

University of Grenoble Alpes — Speech perception, multimodal/multisensory familiarization — GIPSA-lab (CNRS)

Thesis

M.Sc. in Cognitive Science

University of Grenoble Alpes — Machine learning, statistics, and software engineering curriculum

Languages & interests

Languages

  • French — Native
  • English — Fluent

Interests

  • LLM application architecture
  • Agentic systems
  • Brain-computer interfaces
  • Psycholinguistics
  • Developer tooling

Full professional history available on request. Get in touch →