About Me

Romain Vial



I am Data Scientist @ Hyperlex, a SaaS Contract Management Platform driven by AI. Here, I ensure the continuous improvement of the AI engine of our main product. In parallel, I manage several PoC projects to solve our customers’ challenges ranging from NLP to Computer Vision applied to legal document analysis.

Prior to this position, I graduated from MINES ParisTechand from Master MVA (ENS Paris Saclay). This gave me strong technical (computer vision, NLP, statistical learning) and managerial (project management in a research/technical environment) skills.

09/2018: I'm pleased to pursue the adventure with Hyperlex as a full time Data Scientist!

03/2018: I just started my internship with Hyperlex!

02/2018: We had one paper accepted in TIP journal!

09/2017: I will present our paper at ICIP 2017 in Beijing! Feel free to contact me for a coffee!

07/2017: We had one paper accepted to ICCV 2017 in Venice!

06/2017: We had one paper accepted to ECML-PKDD 2017 in Skopje!

05/2017: We had one paper accepted to ICIP 2017 in Beijing!

02/2017: I started my internship with NEC Laboratories Europe in Heidelberg, Germany.

07/2016: I started my internship with Institute for Infocomm Research in the Visual Attention Group in Singapore.

My Career


Development of a NLP pipeline to automate the process of analyzing legal documents for a SaaS contract management platform.

  • Continuous improvement of Hyperlex AI engine in close collaboration with the Product team
  • Managed several PoC projects to solve various NLP tasks (e.g. information extraction, anomaly detection) for our clients

Mar. 2018 - Aug. 2018
Data Scientist


  • Benchmarking state-of-the-art methods for text classification in legal documents
  • Integration of those methods into Hyperlex AI engine which leads to a significant qualitative improvement for our customers

Mar. 2018 - Aug. 2018
Data Scientist Intern

ENS Paris-Saclay

I completed the Master 2 Mathématiques, Vision, Apprentissage (MVA) in parallel of my last year at MINES ParisTech. Master MVA is a prestigious research Master in France, specialized in Computer Vision and Machine Learning. Taught by leading researchers, it aims to bring students abreast with the cutting edge in research.

Attended courses:

  • Computer Vision and Object Recognition (C. Schmid, J. Ponce, J. Sivic)
  • Probabilistic Graphical Models (F. Bach)
  • Reinforcement Learning (A. Lazaric)
  • Convex Optimization (A. d’Aspremont)
  • Algorithms for Speech and NLP (E. Dupoux)
  • Machine Learning with Kernel Methods (JP. Vert, J. Mairal)
  • Curse of dimensionality (S. Mallat)

Sep. 2017 - Jun. 2018
Master 2 Mathematics, Vision, Learning (MVA)

NEC Laboratories Europe

I worked with the Data Science group on memory networks for aspect-based sentiment analysis. We further extended the system to tackle sentiment analysis in financial headlines.

Feb. 2017 - Sep. 2017
Research Assistant

Institute for Infocomm Research

I worked with the Visual Attention Lab on activity recognition in videos where I proposed and developped an object detector based on recurrent neural network.

Jul. 2016 - Jan. 2017
Research Assistant

MINES ParisTech

MINES ParisTech is one of the top and oldest French Engineering School. With more than 1,200 students (Masters and Ph.D) and its 18 research centres covering a wide range of topic, it has built a strong link with companies and industries.

Attended courses:

  • Statistics
  • Machine Learning
  • Probability
  • Optimization
  • Image Processing
  • Stochastic processes

Sep. 2014 - Jun. 2018
M.Sc in Executive Engineering


YoTube: Searching Action Proposal via Recurrent and Static Regression Networks [PDF]
Hongyuan Zhu, Romain Vial, Shijian Lu, Xi Peng, Huazhu Fu, Yonghong Tian, Xianbin Cao. IEEE Transactions on Image Processing, Vol.27, June 2018

Graph Enhanced Memory Networks for Sentiment Analysis [PDF]
Zhao Xu, Romain Vial, Kristian Kersting. ECML-PKDD 2017

TORNADO: A Spatio-Temporal Convolutional Regression Network for Video Action Proposal [PDF]
Hongyuan Zhu, Romain Vial, Shijian Lu. ICCV 2017

Search Video Action Proposal with Recurrent and Static YOLO [PDF] [PPT]
Romain Vial, Hongyuan Zhu, Yonghong Tian, Shijian Lu. ICIP 2017