 Good morning. Good afternoon everybody. My name is Olivier Liman, the president of INT. So basically INT has been involved in OSDU since really the early days and we've really helped you know with the testing of the OSDU platform, the conversion to the CRS conversion package that currently we're helping and validating this CRS conversion with the Geeks testing suite and we pretty much support I mean all the cloud vendors I guess implementation of OSDU. So what we are trying to provide with IVAP is a tool that will really you know support all the exploration data reservoir drilling production CCUS and more you know that that will come through the through the OSDU consortium. So we see our platform is a really you know good tool in terms of data exploration, validation, QC and here what I'm going to show you is a bit of a functionality but we're going to also talk a little bit about the the ability for IVAP to integrate with external machine learning workflow. So INT as a company we're not really you know they're developing a very sophisticated machine learning algorithm but we want a platform I mean to you know for people that have a very unique algorithm to be able to connect them to IVAP and leverage you know all the data selection the data labeling and the visualization of course. So that's what I'm going to show today. Okay so okay so just a very quick you know graphic representation of some of the some of the capability of the you know in terms of the well display the well correlation the core image stratigraphy 3D reservoir production and and so forth. So we we know we do have the visualization for you know pretty much everything that is currently and will be in the OSDU platform. So a bit of architecture because IVAP you know can connect to more than OSDU and we don't make any copies of the data I mean we are you know we have this backend that you see on the on the right side at the bottom and it's just a bunch of microservices with the various data connectors. We also do support real-time and you know in terms of data source so you know just some of the standard database like ppdm pro source open works we even support you know things like snowflake of course or the sequel type of database and so on and of and of course we also support file format like segway open vds open zgy last release and so forth so so very very complete platform and of course you know the IVAP client so the visualization is connecting through the backend using you know REST API WebSocket messaging and stuff like that so so that's a little bit the architecture of the platform and also one thing that I like to it's kind of important so IVAP is a client application but it's also it's also an SDK so if you want to you know pick up some some visualization or dashboards and you have developed your application with Angular or React or something you can you can just integrate the components and and plug it in into OSDU or other environments so so again for the live demo so I'm just going to show a very simple prediction of missing log curves so we're using we're using a TensorFlow deep neural network so I'm going to show ability to search explore the OSDU QC the data build a prediction run the I mean build a model using the prediction and then run the model and QC the result so and and yeah we are writing back into OSDU using the well-bored DDMS so so let me switch let me switch screen here okay right so so this is the this is this is OSDU running in this case in in AWS I mean the version that's available for for testing so what I'm going to do is create a new a new project let's call it OSDU ML and so we're going to bring we're going to bring the now famous TNO TNO dataset what I'm going to do is in addition to you know some of the basic maps we can also connect to our ArcGIS sorry okay and here we're going to let's say we're going to we're going to bring the layer like like with with the fields okay and I'm going to focus so I'm going to just because we we don't have much time so it's going to focus on on on one block of data here but what we're going to do is so we have a lot of search capability here but we're going to we're going to try to find wells that have Gamma, Robi and DT and we're going to we're going to we're going to filter those wells and what we're going to try to do is is predict the neutron porosity so let me select I'm just going to select the the wells around around here and then to my project and then go to the to to the visualization so so so here we have let me take this this first template here so so we have we have our data I can I can select the wells and bring them to a correlation correlation display so again you know this is this is connecting and let's organize everything we see we see all the data together okay and you see here so I have I have the I have the wells so we're looking at the neutron and you see like like those two two wells here I think this this one and this one missing the well actually it's this one or did I forget I think I forgot to add one well to my project let me let me see oh yeah I think I can take this one also here okay so let's add one more here and I'm going to add it to my view okay here we go and just realign everything okay so so then next so again I'm working on very a few few wells but I'm gonna I'm gonna start by labeling them so so so here we're gonna take the the two that are missing the so I'm just gonna call it missing okay and here for for for the purpose of this which is going to select the two the two wells around it to to do our training set I'll put it in blue let's say and also add this one here the training training set and then we'll we'll take the rest to be to be our validation so we'll just call it validation okay so so now what I'm what I'm gonna do is so we have this we have this tool so tab here and and here I have this this curve prediction but the interesting part is the all of those algorithms are not not running or they're not directly connected to IVAP I mean they're just services that are out there in on different server and I'm just calling some end points so the first endpoint is to get the this property panel for the data selection and and the parameter and this panel is actually provided by the service through a JSON form I guess and and so here we have the the ability to select select so for for the training I'm just going to select the the training group we have and then just you use the default logs and here I'm just gonna I'm just gonna select my the input curve so gamma ray dt and ruby and then for the output we'll do the the neutron and I think I can use and for the model so the model I'm create I'm just gonna put this and then just run here okay so as you as you as we run what's happening is the we have a messaging system so the the machine learning workflow I mean is you know sending messages so we can we can also click on this and he'll he'll he'll he'll get get us an update of what the what the machine learning is doing here okay this okay so so you see I mean again I selected just a few a few data sets or maybe the greatest model but for the for for the sake of simplicity I guess and and and speed I'm just gonna leave it like that and then and so what's what's happening now is I can I can run this model and predict the missing curves here so I'm going to go back to this to this machine learning here tab and then we're going to do the propagations for so similar right we have we have the data selector so I'm gonna so for the for the propagation I'm clearly gonna select all all the the groups that have created select the also the default log and then basically so this is another interesting part is I'm going to tag I'm going to tag my data so and we'll see and I can create a log or prediction or SDU let's say for example and I'm going to run this so so again we get the machine learning workflow is just communicating through iVab by sending messages and we can we can look and really you see it's it's it's working on the log it's probably already done but what's happening is we're writing through the through the well DDMS and it takes OSDU a little bit of time to ingest the data back to the elastic search or so we are just really waiting for the elastic search but you start to see our logs you know being being added to to OSDU yeah the prediction OSDU this is the this is the log and while while we're waiting for that I'm going to load another dashboard to to QC the QC the results so I'm going to load this one and and so we will be able to we'll be able to show them to show these are we doing okay and so the the the interesting part here is is I can I can tag the data and you see the you you see the the log that have that have been calculated they have this prediction tag here what we also see okay you don't see double I mean they you know there's some replication of data like the like the tops the some of the logs and everything oops I must have done so here it looks well let's speak okay so of course the the the wells that we have been using for our training you correlate reasonably well but the the the other one the the validation they don't look so good but then that tells us we probably need to to to improve to to improve our model a little bit here and and finally I'm just going to show you the the I guess the result side by side you know into into a correlation view again now we can see we can see the predicted result make everything a bit more compact here and yeah and so you see you you see here again and and so the case k07 dash 03 and dash 01 have have the predicted curve here and and again we can we can really again you see the results very quickly with with the tool but but again I just wanted to point out also in terms of the in terms of the qc is we seem to have copies of of the data but you know by just looking at the information here we'll see that you know one one set of log is the 1.00 version of the schema and the other one is the 1.1.0 but I guess that's an issue with OSDU that we've been reporting that they don't take just the latest version of the schema they you know they keep the old one and the new one so so we you know we are using our platform I guess to you know to to to help to help qc and and do you know various work the various work with OSDU and again I'm not going to show everything but we have you know the ability to show all sorts all sorts of data of course the the seismic the the the tops the faults we are going to release also the schematic the well schematic part of the well delivery and so on and so forth so kind of the wide range of capabilities