Bernardin LIGAN
Data Scientist
Interests
- Computer Vision
- Satellite imagery
- MLOps
- Time Series
- AI for SDGs
- Politics for the development of new technologies in Africa
Contacts
Education
-
2023–2024 Pre-doctoral Student : Hyperspectral Imaging for Analysis of Satellite Unstructured Data -
2021–2023 Master in Modelling and Data Science Mohammed VI Polytechnic University (UM6P), Ben Guerir, Morocco -
2020–2021 Master 1 TIC Institut de Mathématiques et de Science Physique, Dangbo-Bénin -
2017–2020 Bachelor in Mathematics-Informatic Institut de Mathématiques et de Science Physique, Dangbo- Bénin
Industry Experience
Septembre 2023 – Octobre 2023 | Backend GIS-Data Engineer | Valhko
Detailed achievements :
- Identify geospatial data sources
- Design the database architecture with postgis
- Integrate Time Series Data with Timescaleddb
- Design the Backend of the platform with Django Rest Framework and some others django’s GIS modules
- Integrate Swagger Specification (with drf-yasg tool) to automatically generate API’s documentation
- Join all parts of the App by creating services with docker-compose
Mars 2023 – July 2023 | Data Scientist Intern | Center for Remote Sensing Applications (CRSA) : Downscaling remote sensing water-related products to improve groundwater monitoring in arid and semi-arid environments employing multisource satellite-based data
Detailed achievements :
- Collect Satellite data: NDVI, LST, Rainfall, Soilmoisture, TWS
- Preprocess NetCDF images with libraries such as xarray and dask
- Explore the link between the target product and data used as proxies.
- Estimate the target variable as a function of the proxy data using Machine Learning and Deep Learning methods
- Select the best model by using statistical significance tests (Scott-Knott test ) and ranking methods such as Borda count
- Implement a processing scheme for automated remote sensing data downscaling.
- Produce maps of the target variable with an enhanced spatial resolution.
July 2022 – Sept.2022 | Data Scientist Intern | Green Energy Park | Structural Health Monitoring: Development of Deep Learning models for the detection of damage in buildings from Time Series data, collected by sensors
Detailed achievements:
- Collect and prepare Data
- Transform Time Series into images based on methods like Gramian Angular Field, Markov Transition Field, Recurrence Plot
- Build CNN models for each method
- Compare the models and select the best for deployment
Programming Skills
- Data Science/AI : Python, Scikit-learn, Tensorflow, Pytorch, Openvino Toolkit, TensorRT, Pandas, Spark
- Cloud : Azure, Azure Machine Learning, Google Cloud (rudiments)
- Devops : CI/CD, Docker, Git, Github, Github Actions
- Front End : HTML, CSS, Vuejs, Axios, Streamlit, Gradio
- Back End : Django, Django Rest Framework, FastApi
- GIS : Geopandas, xarray, Leafmap, QGIS
Training and Certifications
- Azure Data Scientist Associate, Microsoft, Link
- Tensorflow Developer Certificate, Google, Link
- PyTorch for Deep Learning in 2023: Zero to Mastery, Udemy, Link
- Optimize TensorFlow Models For Deployment with TensorRT, Coursera, Link
- Spécialisation Google IT Automation with Python, Coursera, Link
- Deep Learning Computer Vision CNN OpenCV Yolo SSD GANs, Udemy, Link
- The Complete Guide to Django Rest Framework and Vuejs, Udemy, Link
- Python in Containers, Udemy, Link
Hackathons
- First Prize SAS Hackathon: Creation with my team of the best Fund Raising Amount prediction model using the SAS Viya platform (2021)
Projects
- Creation of a website for the detection and rapid assessment of vehicle damage using AI, for car insurance.
Tools used: Tensorflow-Keras, VGG19, Intel OpenVINO™ Toolkit, Django Rest Framework, Vuejs, Axios
- Deployment of an interpretable diabetes prediction application on Google Cloud.
Tools used: Streamlit, Docker, Google Cloud, Scikit-learn, Xgboost, Lightgbm, SHAP Values - eli5
- Realization of an AI-based application, to help people with reduced mobility to perform basic tasks on a computer from their voice
Tools used: Spacy, Vosk, NER (Name Entity Recognition)
- Bitcoin Price Prediction Using Deep Learning
Tools used: N-BEATS (Neural basis expansion analysis for interpretable time series forecasting), Tensorflow
Participation in Academic Events
-
03 to 05 Nov 2022 AMLD Africa 2022 (Applied Machine Learning Days) Africa, Conference, UM6P Morocco - 26 & 27 Jan 2023 | TechInnov Day, Seminar, UM6P Morocco
Two days with eminent researchers in AI like Yann Lecun (his first visit to Africa).
-Day 1) The future of artificial intelligence according to renowned researchers and seasoned industrialists.
-Day 2) The prospects of machine learning including the challenges related to research, academics, and business intelligence - 23 & 24 May 2023 | African Cities Lab 2023 Summit, Conference, UM6P Morocco / Sèmè City Benin
Several seminars and panels with researchers and decision-makers on topics such as real-time data collection, establishing communication networks to connect devices and people in Africa, data analysis to optimize the management of urban resources, and the creation of basic public services.
- 23 June to 27 July 2023 | Climate Change AI(CCAI), Summer School, Virtual
This virtual program provides an overview of how artificial intelligence (AI) and machine learning (ML) can be used to address climate change. Through lectures and tutorials, we learn how AI/ML is employed across different climate-relevant fields/sectors, discuss important considerations for framing/scoping problems, and gain hands-on practice applying AI/ML to climate-relevant problems.
Club Experience
- Since Nov 2022: Adjunct treasurer of the International Student Club(ISC) of UM6P
- Since Oct 2021: Member of Google Developer Student Club (GDSC) of UM6P