Artificial Intelligence and Data Analytics Program Director

з/п не указана

ООО Арамко Инновейшнз 

Ломоносовский проспект, Москва, микрорайон Ленинские Горы, 1с75В

Вакансия в архиве

Работодатель, вероятно, уже нашел нужного кандидата и больше не принимает отклики на эту вакансию

Показать описание вакансии

Требуемый опыт работы: более 6 лет

Полная занятость, полный день

Aramco Innovations LLC.

Artificial Intelligence and Data Analytics Program

Aramco Research Center – Moscow is offering great job opportunities for innovative data and computer scientists specialized in Artificial Intelligence (AI) and Machine Learning (ML). The oil/gas industry has recently shown a great deal of interest in ML and AI while committing substantial amount of resources for technology development to tackle business problems. The industry is full of interesting applications involving massive volumes of data generated from exploration, development, production, and characterization. Many of these data, either time series, high dimensional images or unstructured operation data, bear resemblance to those from familiar ML applications in other industries. However, they are also associated with specific deep domain knowledge that need to be accounted for. Our goal is to develop ML/AI technologies in the broader geoscience area, drilling and development, production and reservoir management.


Seismic Imaging:

Seismic imaging and inversion is another critical area of processing which, based on large-scale numerical simulation of wave propagation, generate structural image using wavefield cross-correlation or invert elastic bulk properties throughout the subsurface volume. Sometimes it also uses other geophysical data, such as EM, MT and gravity, if available. These are butter-and-bread of seismic processing, analogous to ultrasonic medical imaging but with increased model complexity. How to integrate wave physics and numerical simulation with ML/AI, to achieve improved accuracy with quantifiable uncertainty and at manageable computational complexity remains one of the critical challenges.

Geological Model Building:

Many of the seismic/geological interpretation steps can be considered as machine learning problem where certain structural body (fault, salt, et al), deposition episode, or fluid and lithology types are to be identified from the processed seismic data or image. So this is one of the topic areas where ML/AI methods has seen rapid application and large number of reported results. However, most existing attempts often take a page from computer vision and try to classify elements of the geological hypothesis. How to collectively put the element based work together and develop large-scale composite model classification remains a challenging future direction, especially for large-scale 3D seismic survey with potentially noisy data.

Well Logs and Facies Analysis:

The idea in log-based facie prediction is that by measuring the acoustic and electrical responses as well as the nuclear radiations of the drilled medium, we can infer properties about its rock matrix and fluid content and indirectly relate them to the porosity, permeability or fluid saturation of the rocks. Reported work on log based facie prediction using machine learning methods have increased quickly in recent years. The challenges include relatively modest accuracy obtained so far, and limited understanding of how to best leveraging logging physics with ML/AI models.

Drilling risk mitigation: kick detection:

This is essentially a problem of early warning, detection and classification for anomalous changes in drilling process, similar to anomaly detection in ML/AI with applications in automobile or aircraft control monitoring. The challenges in this case include very limited amount of data available at the surface where the potential cause of kicks can be complex subsurface geology or borehole structure. In addition, it is crucial to have a reliable, timely and sensitive kick detection or prediction system which at the mean time is robust against false alert so that it does not disrupt normal operation and become unusable. This is further complicated by fluctuations incurred during normal operations or other issues not related to fluid fluxes.

Production monitoring: electrical submersible pump (ESP) failure detection:

The challenge in this case is the large number of potential failure causes and failing parts while the available measurement data (pressure, temperature, motor current, et al) can be rather limited in comparison. While there is no shortage of accumulated field measurements, labeled data set with logged failure modes are not necessarily sufficient or adequate for accurate detection and diagnosis, especially given expected normal fluctuations in the measurement data.


an experienced individual in upstream oil and gas industry with more than 12 years of experience with a PhD degree. Successful applicants should have strong track records in pioneering the creation of new upstream technologies that leverage artificial intelligence and data analytics.

Ключевые навыки

upstreamPhDartificial intelligence


Ломоносовский проспект, Москва, микрорайон Ленинские Горы, 1с75В
Показать на карте

Вакансия опубликована 21 апреля 2019 в Москве

Написать сопроводительное письмоПисьмо отправлено

Сопроводительное письмо к отклику

Вакансии из других профобластей