From Documents to Discovery: Advancing Research Collaboration in New Mexico with AI

Thank you. My name is Brittany Van Derf, I'm communication and outreach manager at New Mexico EPSCoR. New Mexico EPSCoR is part of a larger NSF ecosystem that has a mission to enhance the research competitiveness of targeted jurisdictions like New Mexico. I'll be supporting these rockstar women today uh along with my partner in EPSCoR related crime, Mr. Dustin Allen, who will be working behind the scenes to make it sure everything runs smoothly. And if you have questions at any point, um it Oh, this is going to be about an hour-long webinar or workshop webinar session. So everybody knows. Um if you have questions at any point, just unmute yourself. This is We like to have have a dialogue. Um so just feel free to un- mute yourself. But we do ask you remain muted when possible if you're joining us from a loud like a loud environment, okay? And also uh there will be a recording uh and a transcript of this webinar session that will be made available on our our website and um elsewhere online. So pay attention and look for that. We'll We'll send out details after the webinar. Now, I'm going to turn it over to our honored presenters and stop sharing. And let them take over. >> Brysa, do you want to go ahead and share? >> Okay. I'm muted. I was looking where I can unmute myself. >> Okay. And thank you. So I want to welcome everyone also. Um I am Lori. I am former director of the New Mexico Cybersecurity of Excellence. Jessica Trujillo Torres is on with us. She is the coordinator for the Center of Excellence. And she will be working very soon with the new director. Nathan Fisk has been hired and will be starting next month as the new director. Um he's already agreed to help with this project. And so, when I talk about the director of NMCCoE, that will be Nathan Fisk. And you'll see announcements about his um starting in the center in in the coming month. Um But, the other people I want to introduce, Raisa, do you want to turn your camera on if you can? Um, Raisa Islam is still with us. She is a graduate student. She's a PhD student who's about to finish. Um, and she has been instrumental in getting this project up and going. Noah Frye also worked with us for a long time. He was an undergraduate who graduated in December. And the two of them have done most of the work. Let me just be honest, on developing this large language model system for enhancing research collaboration. So, that's what this is all about. It's part of the EPSCoR project, which had three pillars. One was infrastructure for research. Two was enhancing professional development and collaboration opportunities. And number three was the administrative piece that the EPSCoR main office does essentially with just a few people with this huge collaboration system. And so, the the um pilot project that I proposed was to cross over those first two components of EPSCoR to really help us all find collaborators in our own institutions and then across the state to enhance our research productivity and to be able to do more interesting research with better collaboration. Um, for that, Raisa and Noah, as I said, did the majority of the actual work of building this system. Um and then the system administrators at NMCCoE working with Jessica have done support for it. They run the servers for it. Um and have been involved whenever we've had network problems or whatever. Um so we don't currently have the URL for the statewide system. We're waiting to hear back from the lead system administrator on what that URL is going to be. Um and once we have that, we will have the EPSCoR office reach out to everyone with what that URL is. But today, Raisa is going to take you through the operation of the LLM and how you can use this collaboration discovery tool. Raisa, take it away. >> Hello. Good morning, everyone. And thank you for being here today. Um my name is Raisa Islam. I'm working with Dr. Laurie Liggatt at New Mexico Tech. And we call our project the research LLM. Um the title of today's talk is from documents to discovery. And as the name suggests, we are set out to do exactly the same thing. So I will talk more about um the problem and how we are trying to solve that in a bit. First, okay. Here is uh the agenda for today's talk. So Dr. Liggatt already introduced um the team, NMCCoE and EPSCoR. Next, I will be talking about the in-house LLM or IHLLM as we call it. Then we will talk about how uh participating institutions can participate in two ways. They can run their own um in-house LLM or just contributing uh contribute documents to the statewide instance. And then we will show how in-house uh research LLM works. And finally, Jess will be talking about how IN-M CCOE will be support everyone on this project. So, the problem. What problem we are trying to solve in our research project? The problem is who else is working on this. So, um in any research ecosystem, the knowledge that should connect people is fragmented. It lives inside um published papers, technical reports, proposals, and documents that are spread across uh departments and institutions. Finding related work or um the right collaborator means manually reviewing all those papers, and then uh this process is really slow because we have to do it manually, and sometimes it might be incomplete as well. And it does not scale. So, the problem is sharpest in newer, fast-moving research areas where the informal network is uh have not formed yet, and there is no obvious person where we can reach out and ask who is working on this uh field or area, and how can we collaborate with them. Our objective is to transform uh the scattered documents into structured, queryable um knowledge so that um discovery and collaboration becomes easier. So, how do we approach to solve this problem? Our answer is a research discovery LLM. At its core, it's a retrieval augmented language model that indexes published research and lets uh the anyone find potential collaborator collaborators by topic and expertise. The system comes in into two tiers. The first is the in-house LLM or IHLLM. It runs locally at an institution, indexes the published papers of that institution, and answers questions privately. The second one is statewide LLM. For our state, it's New Mexico LLM. Uh this is an aggregated instance hosted by NMCC that combines contributions from participating in institutions. The statewide layer is what makes cross-institutional um discovery possible that we would uh we might not have known otherwise. >> So, I want to comment here before you change slides, Raisa, that notice that we said we're talking about published papers. Um this Originally, we had a discussion about whether unpublished works um and whether restricted access works could be part of this. But, because of the federal requirements on export controls, we did we decided to not do that. So, we're intentionally not including things that could potentially result in an export control violation. We're only using the documents that have been published. Each institution could make other decisions, but we have not implemented the system to divide those access controls by role, specifically because we didn't want the the possibility of exposure for either anything classified, proprietary, or export controlled. So, that's just a comment on decisions that were made in discussions with our Vice President of Research, um and Dr. Lekha Kulan, and it was important to us to to have that discussion with you all so that you know there's reasoning behind that. Go ahead, Raisa. >> So, now I I will explain how the system works. Um under the surface, this is a pipeline that turns raw PDF files into retrieval-ready knowledge, and we have the stages demonstrated in here. Uh so, it begins with um discovery. In the system, we um input the name and email addresses of researchers who are working on the institution, and then based on the that information, we start paper download auto Based on that, the automatic paper download pipeline is started. Sorry. So, when we start the automatic paper download automatic paper download, it tries to find all the papers online, and if we cannot access some papers publicly, then we create a list of that, and then a separate email is sent to each researcher um asking if they can upload their papers in a Google Drive folder. So, um we have our own email service system for this, and when we send out the email, we uh create a very specific Google Drive folder for that uh author only. And we also uh tell them give them the list of the papers that we were not able to download, and also the guideline how to upload the papers. So, we will be discussing more about that um soon in another slide. And on the background, when we download the papers, we start on uh metadata extraction like titles, authors, affiliations, um dates when the paper was published, um abstract of the paper, and keywords. That metadata is normalized against a validated JSON schema, and then we write that into Postgres database. And [clears throat] finally, um the curated metadata and document uh is fed into the large language model for for training. And um after that, the users can ask query questions and our LLM should be able to answer their questions. On the technology side, the service is built on fast API. Then we have used Rec Pipeline Sentence Transformer embeddings and an Ulama hosted model. So, how it works. This slide shows how the pieces fit together across the state. Each participating institution, say institution A, institution B, and others, they can implement their in-house own in-house research LLM and then connect to the state-wide research LLM. Otherwise, if some institution decides that they do not want to implement their own in-house research LLM, they can just upload their paper in the state-wide version and then state-wide research LLM will be processing their papers and the users will be able to ask questions about that institution as well. So, how the local in-house LLM works. How do you make the local in-house LLM work? So, first you have to get the code from GitLab. I will share the links in a slide later. And then you have to set up >> [clears throat] >> the code according to your university. There will be a quick start quick start guide for that, too. And the collection process This is a chain. Sorry. The collection process um happens in um in two ways. First one is the automated download pipeline where uh we download the papers uh online via search engine. And when we cannot download some paper, we ask the faculties. And then after that, we extract information and train our 11. So, here is how we can set up the in-house 11 for any institution. This is a checklist. Uh so that you can have a manageable list of items to update and change to your own specification. First, we have front end and we have Here, we need to change 11 files. Um 13 actually because in components page, we have three files. Next up, we also need to update the credentials and then back end also. We have secrets folder in the back end. We have environment file and then many other. Let's start the project. Um let's start with front end. In front end, uh we will go to source and then we will go to api.js. >> [snorts] >> In here, we can see that we have um base URL which contains the URL for uh New Mexico Tech. We need to change it in here. And then next in uh index.js, we have Okay, login.js. Uh this one we have public index.html. In here, we have the logo. Right now, the logo contains NMT ways logo, so we will need to change that. And then we have another logo in um somewhere. In here. Where's component styles logo? Look in here. So, we will need to change this actual. Uh we have comp- components alt login. In here, it says connect to NMT single sign-on. You should change that. And after [clears throat] that, we have layout alt header. Um Here. So, here the title says NMT research LLM. You should change the words and use your institutional name. Same goes for back um back header. We have NMT research LLM in here, and then landing header, NMT research LLM. You should change those as well. And styles.min.css. In styles.css, we have used NMT's color scheme, so if you want, you can change uh the color scheme as per your institution. So, this is for front end. Next, we have credentials. It's inside the config folder. We have two files, credentials.json and token.json. Credentials.json is the file that is downloaded from Google Cloud Console after configuring OAuth2 for the application. And token.json is the file This file is created after the file first credentials.json file successfully is successful Google authentication flow. It stores the authorization tokens used by the application to maintain access to the Google Drive API. >> Grace, I think the video stopped. Uh it was playing on my end, so I couldn't say. >> The audio stopped on our end. >> The audio stopped? Example. Next step. I don't know how to I go back. >> No, it's back. >> Oh, okay. So, should I uh play from the from which point you didn't uh get the audio? >> Maybe go back like a minute. >> In a minute? The credentials? >> Maybe not a full minute. That um I think you were doing logo stuff. I think that should work. >> Okay. Is the email for meeting? SP.crt This generated file is used to sign and certify communications to the service provider. And then we have SP.key. This generated key file is used to fulfill identity identity requirements for the service provider. You can generate service provider key pair using a basic command shown in this example. Next up we have environment file. Here the environment file I will not show our .env file for security purposes. Here is an example. You can see what variables we need to set from here. We need to set the hot authentication secret service provider public certificate path SP private key path, SP sign authen- authentication request, and then we have to assign IDP metadata URL, IDP metadata path, SP metadata path. These three are required for SSO connection. And after that we have email connection credentials. First one is email back end. Second one is the email host where you will be hosting your email service. service, then email server port, if you are using TLS or not, email host user, email host password. And then finally the default role of the user. So the default role is set to user. Inside the back end file we have private app auth config file. So here we have front end origin, API origin. You need to change these two and set to your local host server. Next up we have auto-download. [clears throat] What? Or auto-download. So, in here as well, we have default from email, which is basically a no-reply email. So, when we send uh some notification email, uh it is sent from the no-reply email. And then, we have a default admin email. So, if someone wants to contact us, they can use this email address. >> [clears throat] >> Next, we have uh parent folder ID. So, I was going to talk about it later, but uh since we are here, the parent folder ID is for the Google Drive folder where we will be giving access to the other researchers where they can upload their research files. So, this is just the folder ID of that drive. And inside this drive, we will be uh we will be creating subdirectories. And then, we have email services. So, here we have a uh function called send generic email. It has everything uh according to NMT's policy and what we are trying to send. We have send email. This is also how we want to do it. So, you should want to change the lines in here in the generic email and uh notification email. Up next, we have doc handler. In doc handler, same to the previous file, we have default from email, default admin email. Next up, we have helper. In helper, we have Google Drive services. So, in here, we have to set up the config directory. So, this is basically the folder path where the uh credentials are. So, in our case, we already showed that uh the credentials are inside the config file. So, we are just saying config. You can name it anything else as well. Or good. Next, we have institution name.json. So, this is a JSON file containing a list of um names that your institution is popularly known is um known as. So, for our case, New Mexico Tech is also known as uh the uh NMT NMTech, New Mexico Institute of Mining and Technology. So, what other abbreviations you you have, you just you can just put it in here. And then, this file is used in two other fo- um two other files. So, we're using this in auto download. Here, we are reading from JSON, and it uh just copies everything from institution name.json. Also, we use it in detail extraction. Here. Next, we have main.py. So, here, you might uh you might want to change the title. It says equal element API. If you want to change the name, the description, and then middleware. The middleware is research element cybersecurity uh dot nmt.edu. You should be able to change it to your institutional uh URL. So, that's pretty much it for front end, back end, and configuration. Next uh is the instruction about running the application. Uh so, in order to run the application, once uh you have all of that configured all of this configured, you will use podman compose was up. Build uh sorry, podman compose up build uh if you are using podman, otherwise it can be docker compose. We have used podman because uh it adds more security. The YML file is configured uh this initiates the starts of all volumes and rebuilds with any back end updates. So, if you update the back end files, this will update with those changes. Up will uh simply bring back up the last run occurrence of the application without any new updates. If uh you simply use compose up, you have to use um build to build with uh new changes. Then compose down, we have compose down in here. Uh it will exit the process gracefully, but um then compose down minus V exits and clears all volumes. This is helpful if you want to check and tweak databases and models. This does get rid of everything in the database. So, keep that in mind. Podman system reset. So, we have it in here. Podman system reset helps clear what Docker intends to load within the application. This is just stored image files, stored volume files from past runs, but sometimes it's helpful to reset this. It is also recommended to reset before your final run whenever you are intending to finally run the application. Next, we have a guideline for the users, how they can create an account in Research LMM, how to log in, and how they can chat. >> Now, this is a simple landing page with a simple introduction to the app, as long as simple ways to use it, about who we are, about our mission, and some helpful links at the bottom in the footer. These links take you to the contact page where you can input your email, subject, message, and we'll get back to you very shortly. The FAQ page, which holds information about the application. >> [clears throat] >> Right now this is basic uh introduction to the application back end and how things are working as well as um how you can use the app to find collaborators. This will update as we receive more questions and more feedback, but as it's used right now it's just like a introductory guide. Um, the help center is yet another contact page where you put your email, subject, and message, but this is dedicated a little bit more towards help and support so that we know that you need aid in how to use the application. Legally we have the privacy policy, the terms of service, and the cookie policy. You can read these and by logging into the application you agree to these implicitly. So, we'll log in. And by logging in we utilize single sign on. Single sign on through normally your university or institution. This means that I as a New Mexico Tech person am able to log in with my New Mexico Tech credentials uh for safe, secure, efficient, and very fast um login. So, this will be set up with your university or institution's credentials and you'll be able to log in with those. So, I will log in now. And this takes us to the basic dashboard page. At the dashboard page it we see a chat page. The chat page is rather simple asking that the user simply input a message to the text field with buttons to send and buttons on the header to check your account and log out of the application. To find papers we can simply ask the chatbot what papers do you have on machine learning? So, I'll take that. Send this back over but not without first saying hello to the application. So, hi there. What papers do you have on machine learning? And now we see a nice list of papers along with their authors. So, this is So, yeah. Let's say that we believe that paper three may have some information that we need. So, let's say that we believe that this is proper, this is good. Let's start finding more information about these authors, right? Can you Oops. Find the authors in paper three. And now, we see papers that match our authors, provide contacts. This is William Ampomah. And this allows us to find more papers from them. So, let's say that we're particularly interested now in paper five. So, now we will ask the application to provide a basic abstract of the paper. And this will help us gain knowledge about what is inside. So, we notice that William Ampomah, previously, has fields inside of New Mexico Tech. So, let's see what papers do you have from authors at New Mexico Tech. And we go through, we find the authors that work at New Mexico Tech. In this case, Anthony Morgan for this paper. And the paper. Now, we could ask more information about this or etc. The user page however, can be used to further guide this prompt. We'll see that here. We'll go to the account page. And this is the basic information about you, the user. So, you can see that during the onboarding step I have logged in with my display name. This is the email that was brought back from SSO. The time zone that was brought back from SSO. I assigned the organization that I'm a part of. I also assigned whether or not I want research connections, the areas of research that I work in, and the role inside of the application that I am. This will just show as user for you. You can edit this to state that you want connections for research or not, and keywords to guide the prompt. So, by checking this box, you agree that based upon your areas of research, you would like to be contacted in those fields for more research, right? So, by clicking this box and stating my area of research is machine learning, now this is a simple landing page. Papers related to reinforcement learning. What we do is we attach the context which >> Um so, you have uh watched how a user can create their accounts and then also check with the research LLM. Next up, we will uh see how what are the features that an admin can access, how they can upload uh new papers to uh train the to enhance the research LLM, and other features. This tutorial aims to introduce admin users to the to their role requirements as well as detail how the admin users are able to perform those functions within the application. The research LLM is intended to make the system super easy and secure on any system. If at the end of this tutorial, you have any question, please contact us using the contact page or by contacting an MCC OE. Application introduction. This application at face value is a rag stain LLM application built on top of a JMA 34B model. This uses simple embedding cross encoding and adapting models from hugging face in order to train documents and be able to pull using queries. Each of these hugging face models are trained using the past documents. This means that each model in incorporates each of the information of the file and places each of them on same graph. Similar papers or words are placed closer together and the they will be pulled together with similar queries. In this way, whenever a user um input inputs a prompt, that prompt is embedded using the same input or that we trained using our data. That way, whenever we take out take our prompt, we can place it near similar papers using similar wording. So, prompts provide um similar papers and this is how it works. Admin user privileges and roles. So, an admin user is responsible for ensuring the functionality of the application by performing actions that update, delete, and um retain the model. They are also responsible for handling the user database uh where they will have the sole responsibility for promoting and demoting users to the admin role, as well as deleting users when requested. Admin users are given access to a special version of the chat page that will allow them to enter the admin control page where they will be given uh many more options to manipulate the application. So, initial login as um an admin. Admin privileges are assigned from the university or institution. If you are an admin, congratulations. Now that you have been uh given admin privileges, you are able to access this admin controls page where you will be able to make sure that the model is updated, read, and retrained with the most accurate documents from the databases at the most reasonable times. And you can download new PDFs using a CSV file containing the name of researchers and their emails. Uh you can also upload new documents to be used for training, update the user database, and perform the training loop, which is uh preprocessing and retraining. Other functions of the application like chatting, um the account, logging in, logging out, and um checking the footers or anything like that, they work just as normal just with this added permission. Let's get started with the demo. First off, we need to log in. Here, if we go to the URL, here we will have a login option. If you click on it, you will uh see another option, sign in to SSO. Here, we have uh landed on admin page. Uh here, you can see that we have admin controls option. So, if we go there, you can see we have many other options. First, we can check our user account. Here, you see I am registered as Faiza Islam. This is my email address. I can edit and then edit the areas of my research. I can uh choose to be connected or contacted with potential connections for research, or I can opt out. I have both the option, and if I am choosing it, let's save this. Let's go back to dashboard. Now, in here, we have multiple options. Uh Uh, one is upload new documents for training. In here, we can upload a single PDF or a zip file, which, uh, will contain multiple other PDFs. Let's see how this works. Here is the interface for uploaded document. If we click on choose a file, we will go Let's do this. So, I have paper one in here. I am choosing this. >> [snorts] >> And then when I click on upload, it will take some time depending on the size of the file. If you are choosing to upload a zip file, it might take longer. Now that I have uploaded the, uh, document, let's go back to admin dashboard. After I have uploaded a file, we need to pre-process. And then, after pre-processing, we will, uh, follow up with free training model. Next, we have download new PDFs. So, this option gives us, uh, a automated pipeline that will automatically download research papers given a CSV file containing the name and email addresses of, uh, the researchers of your institution. So, here we can choose the file. Let's choose it. I have tester group CSV. I will upload this one and then upload. The file has uploaded successfully. And after the file has been uploaded, it will run in the background. It will use that CSV files to automatic automatically download papers. And when it is done with downloading all the papers, it will give me notification in here. So here I have previous notifications where I have automatically downloaded papers. I have pre-trained model and you can do many more things. Along with notification in the notification bell icon, you will be also be sent an email. Now here we have handled the database. Here we can see the list of users, purge data, promote a user to admin, de- demote a admin to a normal user and delete user account. Let's see our list of users in here. So right now I I am the only user. Let's go back to dashboard. Next off, we have set time. Set time allows to set a um definite time for retry downloading. So I already mentioned that we when we upload a CSV file, it automatically downloads file, but it is possible that when it is trying to download files, it may find some files publicly not accessible. So it will keep a list of those files and then periodically try to again download them. So we have received a notification in here. You can see Let's see. So it says auto download finished. So we can now retrain model. It will also run on the background. Next we have copy files to drive. It takes input of um the input in here is a link of a Google Drive file. So, when we put the link and then submit, it will automatically collect all the local files which we have been uploaded using upload new documents or we have downloaded using download new PDF. And then it will copy all of them into that specific Google Drive. This feature helps to um use all the in-house LLM PDFs to use for training in state-wise LLM. We have got another notification. So, it says model retraining is also done. Um then we have list JSON data files. Here we have the list of files. We can hide JSON data files. Let's go back to the um dashboard. So, in here we can just ask questions. Let's see um give me name of a computer I need research. So, it has given us a list of researchers. Here we can see Dr. Laurie Lee Brock. And then we have the option to log out as well. >> So, in this uh video we will see we are we saw how an admin will manage the in-house LLM. Next, as we mentioned in our um automated download pipeline, when we fail to access some paper publicly, we send email to the author of that paper. So, uh in that instruction or in that email, they will receive a list of papers that we were not able to download, and then we will send them a guideline for upload, like what to do and what not to do. So, the paper they will upload should be searchable PDFs, and the listed ones. If they want, they can um also upload their unpublished work, but we recommend not uploading proprietary um work which is not which should not be accessible by other people, because right now we do not have role-based access. And they should not um upload drafts or incomplete manuscripts or presentation files, um any file that is that has sensitive or confidential um data or restricted access. And how to upload? In the email, we will send them a Google Drive link, which is dedicated for that specific researcher. No one else can um access that uh Drive folder. They should sign in using their institutional email address, and then they can simply drag and drop uh drag and drop the PDF file or select to upload file, and they should wait until the file upload is finished. We suggest not to upload duplicate files, and then upload only if there is any already existing files from their previous uploads. We encourage not to delete them. And if they have any issue in the email, we will send them share the um contact information of an admin, so they will be able to contact the admin. So, um how we can contribute documents directly to statewide LLM. So far, we have um discussed how to implement in-house LLM, and then in in-house LLM, showed that we have a option where we can uh copy all our local files just in a click from a Google Google Drive link. And if we want to uh if any institution decides that they do not want to implement their own in-house LLM, they can have one Google Drive folder where they can collect all their searchable PDFs, bundle them, and then we can use that upload document file upload document option, and then that way all their research documents will be uploaded in the statewide version. Now, getting admin access, um I would request Dr. Lorelei Brock to uh >> [sighs and gasps] >> share more about how someone can get admin access. >> So, our um vice president of research has had um numerous discussions on this. And um the decision's been made that we will ask that you contact the new director at NMCCoE, Nathan Fisk. You can do that through the email address that will be shared at the end, that is nmccoe@nmt.edu. And there you'll ask for admin access to the statewide system. That'll be one person that represents each research institution. And that um access should come through your vice president of research or whoever your research lead is. Um Dr. Fisk will be verifying that person is the correct person through your research office or your um institutional leadership. And then he will use that to provide access through the sysadmin team and support at NMCCoE. At the moment, um since he doesn't start for about a month, um that would be through Jessica Trujillo Torres, and I will help her as needed with that. But Dr. Fisk will be the the contact for this system, the lead contact for this system in the long-term. Go ahead, Raisa. >> So, up here we have uh some uh important links. So, you our Git is um our code base is hosted in git.cybersecurity.nmt.edu. And the project is named equal uh NM real LLM. So, if you want, you can download the code from here. I believe this should be accessible publicly accessible. But, if not, you can contact NMCC OE office. You can contact Jess, and they will be able to help you. And then, we have the quick setup guide, the PDF version in this Google Drive link. We have user tutorial for both in-house LLM and state-wide LLM. Then, we have admin user guide for both again, both for in-house LLM state-wide one. And our NMT research LLM is hosted in this address. It should not be right now. It should not be uh able to access publicly, but um the state-wide one is currently on progress. And when that is done, you should be hearing from EPSCoR or Dr. Lorelei Brock or Jessica. They might make the announcement. And our YouTube videos are available at this YouTube channel. We call it NM research collaborations. So, if you go to that channel, you will be able to see the YouTube videos. Yeah, that's pretty much it. If you cannot access any of these, please reach out. Someone might be able to help. I would ask Jessica to say something about NMCC OE, how they might support. Is she here yet? >> Yes, she's here. >> Okay. >> Jess, you want to come on? >> Hey, yes. So, um >> Your video's not on. It's the camera's covered, I think. >> There you go. >> There we go. >> Okay. Yeah, so NMCC OE is here to help Raisa with all her projects that she's doing. We are also helping her with the servers here at the center. Like they mentioned, we have our system administrators that are working hard with her and try to make sure that she keeps everything going smoothly. We're here to help her with onboarding. We're here to help her with traveling. We're here to support her in any way we're able to and we help a lot with the with, I guess you could say the background of her project. We're not seen, we're just known. >> And and in the future NMCC OE and the new director is going to be the host for the statewide system. So the NMCC OE will be continuing to support that. Um as long as that server is able to do so, which we purchased a server specifically for this and it's housed in NMCC OE um and therefore the director the new director Nathan Fisk is going to be the point of contact for that once he starts next month. >> Yes, and then once he starts next month, we will also introduce him to everybody and maybe have a short meeting or something so that he can get involved more and he could get a little bit more details on what was going on and he could ask his questions as needed. >> So we look forward to introducing him to everyone. Do you have anything to wrap up, Raisa? Does anyone have questions? Here I am. I guess Raisa If you have questions, just unmute. If you do have questions, just unmute yourself. >> Please. Otherwise, I want to thank Raisa for all of her work on this project. She's about to graduate and go away and do great things as a new doctor. Um so, thank you very much, Raisa. And I guess we'll turn it back to you. >> Um all right. Awesome. Once again, thank you to all of our presenters. Thank you, Raisa. Thank you, Dr. uh LeBrock. And thank you, Jessica. As a reminder, we're going to have a transcript and recording of this on our website and also on YouTube. And we thank you so much for joining us today. Have a great day and thanks for joining us. >> Thank you. Have a good day. >> Bye, everyone.