Criteo

•	Designed, and developed an autoencoder-based approach to the 12 bits PSB limitations
•	Demonstrated the robustness of bidding models trained on noisy traffic, as imposed by the PSB 
•	Designed data quality checks, and pipelines, specific to PSB traffic

Machine Learning Engineer

NASA Lifelines

- Acted as a voluntary scientific consultant for an NGO interested in monitoring agricultural impacts of ongoing conflicts in a Middle Eastern region.
- Provided a MVP code for road pratictability in GEE to an NGO interested in mapping ongoing road accessibility in conflict regions.
 - Participated as a satellite imagery expert to a 3-days humanitarian simulation held in Geneva. 

The Earth Science Review Board is a group of experts in the field of Earth Sciences and Remote Sensing collaborating on advancing Earth observation technologies for impactful humanitarian efforts.

The work of experts is voluntary, and involve punctual support of humanitarian non-profits with Earth Observation expertise and consultancy.

Expert member of the Earth Science Review Board

ONERA - The French Aerospace Lab

Pursued my Ph.D. education part-time at SONDRA and part-time at ONERA (in Information Processing Department, within the Image, Vision, Learning research team). More details below...

Ph.D. student in Deep Learning for Radar Satellite Imagery at SONDRA Laboratory

CentraleSupélec

Working on problematics of change detection in SAR Time Series of forests with the help of Deep Learning methods under the supervision of Régis Guinvarc'h, Laetitia Thirion-Lefevre & Elise Colin.

SONDRA Laboratory is a laboratory mixing 4 entities: French ONERA, Supélec (known today as  CentraleSupélec), the National University of Singapore and the DSO of Singapore.

Summary: SAR time series are becoming more accessible thanks to the recent big data context in Earth Observation. They are valuable for change detection. Recently, time-processing algorithms can be used to detect anthropic activity between two dates, urban expansion, or maritime traffic, for example. However, the understanding of the SAR signal temporal behavior is still partial and it may strongly impact the thresholding step. In particular, recent studies using time series over agricultural or ground areas have highlighted chaotic temporal profiles. For these environments - and maybe others - the speckle is actually not stable, presumable due to radiometric variation following the variability of weather conditions. In order to better understand the nature time-variations and to better take them into account for efficient algorithms, we propose a temporal analysis of these profiles, by using deep learning techniques applied to large scale open-source data (Sentinel 1, UAVSAR piles, TerraSAR-X archive). The possible contribution of the interferometric coherence information in the robustness of the detection will also be studied.

PhD student in Deep Learning for Radar Satellite Imagery at SONDRA Laboratory

Teaching assistant involved in various EO-related teaching activites with CentraleSupelec students.

1st PhD Year teaching (2020-2021):
 - monitored and advised two groups of 4 students researching the potential of SAR multitemporal imagery for Canadian Forest monitoring, for a week-long project (~ about 27h total)
 - co-managed two groups of students working on road segmentation from optical and/or SAR imagery using Machine Learning techniques, for a year-long project (~ about 10h total)

2nd PhD Year teaching (2021-2022):
 - Co-supervision of pratical sessions of SAR Image processing, using SNAP software for various tasks, including polarimetric image analysis or change detection (~ about 10h total)
 - Co-managed a group of students working on improving road segmentation from SAR Imagery using Machine Learning techniques, for a semester-long project (~ about 10h total)
 - monitored and advised two groups of 4 students researching the potential of SAR multitemporal imagery for Canadian Forest monitoring, for a week-long project (~ about 27h total)
 - co-supervision of a group of 4 students for a research project on evaluating the burnt condition of canadian forests using SAR, for a semester-long project (~ about 20h total)
 - co-supervision of an intern on the topics of agricultural modelling using SAR and optical satellite imagery. Supervision involved the monitoring of his progress, and guidance for the directions of his work. His internship lasted approx. 5 months

Teaching Assistant

European Space Agency - ESA

As a visiting researcher at Φ-lab, ESA-ESRIN, I pursue my Ph.D. research in change detection and unsupervised learning of SAR time series of forests.

Visiting Researcher

e.fundamentals

Industrial Placement as part of the DataLab MSc program that led to the redaction of my Master Thesis.

 - Use of multimodal machine learning algorithms for duplicate product identification in a multi-retailer database.
 - Development of multimodal siamese networks with different losses evaluated and benchmarked (contrastive loss, triplet loss).
 - Analysis of sampling strategy: mini-batch hard sampling, semi-hard sampling, uniform-random sampling, mini-batch distance weighted negative sampling
 - Comparison with ML approach (PCA + Decision Tree + Handcoded textual features)

Machine Learning and Computer Vision Engineer Intern

Thales

- Built Deep ConvNet model to perform building segmentation in satellite Imagery
 - Benchmarked SOTA architectures such as Mask-RCNN, Deep UNET and its variations
 - Written code in Tensorflow/Keras + provided a deployable Flask application to run training, inferences, to monitor their progress as well as to keep statistics on the training (other features included: building datasets from uploading images and shapefiles to the application, adding custom models to the application)

Weekly to daily reports/presentation were provided to the team. Therefore, I have had to summarize my work in order to make it understandable for almost anyone. This was a very gratifying internship and I want to thank the team DBGS-IG of Thales AVS Osny for its workplace quality.

Deep Learning Researcher Intern

Cap'EISTI

- Secretary of a french Junior student association. 
- Acted as the proxy between the school and companies looking for students to develop their projects (e.g. web dev, software envineering..)

Secretary

Freelance | Self Employed

On the side of my engineering studies, I did a couple of freelance missions:
 - Added new website functionalities (account management, product management) to a WordPress website
 - Created a WordPress acting as a showcasing website for a small massage business
 - Created a React Native app for a Start-up project

Freelance Software Developer

Worldline Global

- Development of a RESTful architecture for a back-office managing electronic transactions using Angular, Spring Boot, COBOL programming languages alongside a z/Os running system;
 - Built a modern code while considering industrial hardware restrictions;
 - Followed a waterfall software development scheme.

Fullstack Software Engineering Intern

🌐  
https://dimartinot.github.io
🌐

I am a Ph.D. student in remote sensing deep learning with a passion for Earth Observation & Deep Learning. I have built multiple open source projects exploring EO data from a different perspective (to build learning algorithms to track land cover on time series of satellite imagery or simply to explore a dataset of meteorite landings records).

Trying my best to stay competitive, I arrived 1st at the Sentinel Hub Script competition of 2020 as well as 3rd at the Data Fusion Contest of 2021.

With a strong software engineering background (Python, Java, C, C++) from French engineering school EISTI, I now find myself writing Medium posts about Machine Learning & Remote Sensing projects that passionate me and make me want to share my knowledge and passion with others.

Skills: Machine Learning, Deep Learning, Computer Vision, Remote Sensing, SAR

Taplio

Thomas Di Martino's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Thomas Di Martino's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Thomas Di Martino's Linkedin Analytics