alexandre.

ABOUT · PARIS · INDEPENDENT DATA & AI CONSULTANT

Alexandre Dahan, independent Data & AI consultant.

I learned rigour in a field where AI has no room for approximation: healthcare, FDA, imaging. Today I apply it to your teams’ GenAI agents.

I design and productionise GenAI agents that actually work with your teams: legal, sales, healthcare. On Google Cloud, with measurable guardrails.

Book 30 minutes See the projects

Available now

Alexandre Dahan

2017 → 2026 · Paris · remote or on‑site (Europe)

CAREER

Nine years, from medical imaging to agents in production.

Each step added a new requirement: standards, clinical evidence, data, real time, then taking agents to production.

  1. 2017 – 2018

    GE Healthcare

    Software developer

    C++ DICOM SSO reader, standard-compliant dataset generator and real-time 3D rendering with VTK: POC validated, groundwork for production.

  2. 2019 – 2020

    Physics for Medicine (Inserm / ESPCI)

    Research engineer

    Automatic segmentation of the aorta and left ventricle with deep learning: published in Frontiers in Physiology, 2023.

  3. 2020 – 2021

    Techsomed

    Business development / FDA liaison

    FDA De Novo granted, clinical trials at the Mayo Clinic and a €9.5M co-development proposal with Servier.

  4. 2022 – 2023

    Cegedim

    Innovation project manager, health data

    FHIR health data warehouse, ETL pipelines and medical ontologies: POC validated and handed over to the teams.

  5. 2023 – 2024

    Echo AI

    ML / DL engineer

    Multimodal voice + text model to classify emergency calls into 5 levels, with a stroke / heart attack risk score.

  6. 2024 – 2025

    Devoteam

    AI engineer

    Multimodal RAG / KAG on a large corpus, annotation platform and reproducible evaluation protocols in CI/CD.

  7. 2026

    Converteo, Google Cloud partner

    Data & AI consultant, generative AI

    Three GenAI agents taken to production on Vertex AI and ADK, used every day by lawyers, managers and sales teams.

  8. October 2026 →

    Independent

    Data & AI consultant

    I put that rigour to work for your teams: GenAI agents that go from POC to real use.

    Today

Education

  • 2024 – 2025 Machine Learning Engineer Datascientest MLOps, Kubernetes, TensorFlow.
  • 2015 – 2020 Engineering degree (Diplôme d’ingénieur) ECE Paris Health & Technology major, Data Science track.

EXPERTISE

Four families, one common thread: proof.

GenAI & agents

Designing agents that abstain rather than invent.

Vertex AI · Gemini · ADK · Agent Engine · RAG / KAG · Prompt engineering · LLM-as-judge · Agent memory · Anti-hallucination guardrails

LLMOps & productionisation

Deploy, trace, secure and maintain on Google Cloud.

FastAPI · React / Next.js · Cloud Run · Firestore · KMS · OIDC · CI/CD (GitHub Actions, Jenkins) · Docker · Kubernetes · Langfuse · Cloud Trace · Cloud Logging · MLflow · Airflow · ZenML · BentoML · Prometheus / Grafana

ML / Deep Learning

Vision, signal, text: models trained to run in real time.

Python · PyTorch · CUDA · PySpark · SQL · TensorFlow / Keras · Computer vision · Signal processing · NLP · STT · NER · W&B · DVC · TypeScript · C / C++ · Java

Healthcare & regulatory

Standards, regulatory submissions and clinical trials.

DICOM · FHIR · Medical imaging · Medical ontologies · FDA De Novo · CE marking · KOL and clinical trial coordination

Certifications

  • 2026 Professional Machine Learning Engineer — Google Cloud
  • 2026 Generative AI Leader — Google Cloud

Languages

  • French (native)
  • English (C1)
  • Italian (B2)

Based in

Paris · remote or on‑site (Europe)

HOW I WORK

Four rules I don’t negotiate.

  • No dead-end demos

    A prototype only counts if it has named users and a production date.

  • Maker ≠ checker

    The agent that produces is never the one that verifies: a sceptic reviews, and programmatic verification has the final say.

  • Every LLM call is traced

    Prompt, tool called, cost, latency, judge verdict: everything is in Langfuse and Cloud Trace.

  • The business decides, the agent executes

    Compliance rules are coded deterministically, preferences come from senior staff, the agent applies them.

An agent to take to production? Let’s talk.

Let’s talk [email protected]