Bruno Siarry

Data scientist in the Data Marketing and Data Capabilities squads (transversal teams of 5-10 persons). Stack : Python, GCP (BigQuery, VertexAI, ...), Airflow+dbt, MLflow, Jenkins, Github+Jira, Tableau In the Data Capabilities squad I am focused on fraud fighting and moderation: - Improving our MLops stack to prevent frauds of several types (fake trips, bookings and payments). - Monitoring and reacting to distinct fraud incidents along with our stakeholders in the fraud guild. - Implementing the DS training and evaluation pipeline for Sphinx, our text-moderation program, based on LLM embeddings. In the Data Marketing squad I was focused on proving data driven insights to improve marketing performances. Main projects : - Marketing Mix Modelling (MMM) : measuring the impact of our various marketing channels - Search campaign optimising : modeling timeseries and seasonality - Conversion Lift: A/B Testing of marketing campaigns (in partnership with Meta and Google) I also collaborated to the DS Chapter : organizing MLops talks with various companies of the parisian Tech environment, contributing to our Tech watch and trying new tools on side-projects.

As a consultant, I realised various Data science projects, from POC to production ready ML. Projects include: - For an Electricity group : putting into production the backend of an NLP solution for text summarising and classification. Stack: Python, Flask(+FlaskWTF, flask-session, CORS+CSRF), BERT, Gitlab CI/CD + k8s - For a Luxury brand : demonstrating the feasibility of high precision digital sizing, for online tryout of jewellery. Stack: Python, OpenCV, CNN - For an industry company : creating a simple interface for accelerating industrial plan reading and classification into a catalog. Stack: Python, OCR (Tesseract), Streamlit, DecisionTrees

Creation of ML solutions specific to the clients needs, POC and in production. Treatment of industrial and meteorological data, as well as NLP. Preparation and presentation of Data analysis and visualizations to the client, to generate insights. Improvements of the Axionable Methodology for Reliable AI (1st company certified by the LNE in France), with a focus on : biases, explicability, robustness and AI project good practices. Main projets included : - For the energy sector : creation of a software (in production) using a tailor-made ML classification algorithm, to assist the engineering department in the modernization of an industrial facility. - For a big insurer : analysis of climatic data to asses and predict losses due to natural disasters. - For internal development : enhancement of the Axionable Delivery Method (ADM) framework, to prepare for its certification. Preparation of the audit of our processes by the LNE.

Development of novel super-resolved microscopy setups, image analysis techniques and experimental protocols. Construction of STED and STORM setups and acquisition of fluorescence microscopy images. Quantitative analysis (using Python and Matlab). Main results: - Development of SIMPLER: novel technique for 3D nanoscopy with <10nm resolution (to be published in Nature Comm.) - Observation of super-resolved FRET with STED setup (to be published) - Design of image analysis post-processing and analysis pipelines using Python and Matlab (clustering and colocalization).

Teaching in physics at Bachelor level. 192 hours over three years. Optics, electronics, acoustics, ...

Development and setup of pulsed lasers optical experiments, for the study of nanostructures (In/As Quantum Dots). Spin dynamics modeling (via semi-classical models) using Mathematica. Experimental techniques: He cryogenics, continuous and pulsed lasers. Main results : - Setup of a novel detection mechanism for pump-probe experiments, using heterodyne amplification, leading to an 1000-fold improvement of setup sensitivity - Manipulation of quantum dots spin coherence via the application of a magnetic field - Evaluation of the the coherence time of hole spin in In/As quantum dots via Mode-Locking

Design of MEB image analysis algorithms for characterization of nanostructures (photonic crystals) with MATLAB.