Transcripción automática, sin corregir a mano. Puede tener errores.

00:28 e e sir

01:58 everyone uh it's time for the second

03:13 keynote of today and let me introduce you paa Vel um paa Vel is a data scientist and full stack developer with uh over 23 years of international experience leading multidisciplinary teams in the private public and nonprofit sectors uh she served as coordinator of repositories research and foresight at the National Council for Science and Technology in Mexico where she helped to coordinate the covid-19 efforts for her work in the field of public interest data science she has been recognized with the MIT innovators under 35 latam award named one of the BBC's 100 inspiring women and received a fellowship at the Bergman Client Center for inter of no for internet and Society at Harvard University so muchas thank you for having me and uh welcome to Mexico who like to all of you that are not from here I would normally

04:14 like to um to do this in Spanish but because this is an International Conference I was asked to do it in English but so I troduced uh my um presentation so we understand understandable so as uro mentioned my name is paa Vel uh and from 2018 to 2021 um I had a privilege of uh leading uh data science team under the uh National Council for Science and Technology in Mexico's government um and in this team we basically were asked to provide a data science muscle from uh for both the the federal government but also uh articulating with uh the universities and the research centers so um this was a very very interesting um opportunity and I'll give you a little bit of context this so

05:18 kit um is the National Council for Science and Technology um is the head of the science technology and innov and Innovation sector in the government of Mexico and uh it basically incentivizes uh scientific research uh through many means right like for instance uh research projects or um stiens or many other like uh like fellowships or any like monetary incentives for U the development of Science in Mexico um it also coordinates 26 research centers um they are like they cover over most of the science uh branches uh and we work with them uh to basically create uh impact using their work right like uh what I saw when I first came to gasit was that a lot of uh like a lot of very interesting and very

06:20 sexy um science was being uh performed in all these research centers but they didn't have much um impact on you know like uh actual problems right like it was mostly done to publish papers and to collaborate with other researchers uh so they didn't have a an applied Avenue um so that's basically what we uh what we wanted to do basically um build the bridge between these amazing scientists and the federal government and all these uh the the Ministries uh in charge of for instance uh security or health or Transportation uh whatever like if if we found a project that uh could be interesting or to be applied in any of these um uh Ministries we will definitely go and um yeah like build the

07:21 bridge and manage the the stakeholders and everything so yeah it was um very very uh interesting to me uh and when we um found uh like a methodology to enable these collaborations we called it the ecos naal informatico which basically means the national informatics ecosystem it's a very weird name uh but uh it's basically just what what what I just described right like articulating between data science or data driven projects uh performed in the research centers and the federal government and we have some access um so I'll give you some examples uh the first um any or ecosal informatico that we created was um about mexicos disappeared

08:24 and missing persons right so in the first days of government um um Mexico's president said that the top priority was finding the missing persons like the Disappeared right and as many of you know Mexico now has more than 100,000 disappeared people and this is a big big tragedy uh in Mexico and the government recognized that and basically asked us to um enable collaboration uh between scientists that have the technical power power um and the commission National De Bu which is the how you say like the the commission it for searching this disappear people um so this was very very fruitful we created some projects um for instance oh we created um one project that was

09:28 very impactful um to deal with the the Dirty War which is Mexico didn't have a um a dictator dictatorship but we had a single um a single party uh in power for over 70 years so it was like the perfect um dictatorship right um in and in the during the 70s uh like starting from the mid 60s to the end of the 70s uh Mexico's Army basically U disappeared activists and students and uh any type of descent right like and they did that um and what's very strange is that they documented everything right everything is in um like metric tones tons of uh

10:30 paper and boxes and it's in a warehouse and it's very like it used to be like super super confidential but in this government it was decided that it should be open at least for researchers and eventually it was going to be open to the public um so the national um had access to these documents and but they didn't have the means to make sense of it right it's like yeah metric tons of paper um so we helped develop at least the road map to make sense of these documents by uh starting for like um digitalization but also um recognizing the text uh finding logos and signatures recognizing names and making uh repository with uh ontological search that Ena The Searchers to make sense of who was where right so for

11:33 instance uh we could know that if a person was disappeared in kakan Sinaloa uh and the the general Cel no like the um headquarters for the Army was in kakan and it was led by this General we knew uh which times um they like coincided and uh made a match so and so they could basically connect um their yeah the case with other cases um using this this search engine that was created um then we also and this this was one of the most uh like the one of the hardest experiences of my tenure in kit uh which was the search for contemporary missing people or disappeared people basically the

12:33 consequence of the the war of against drugs since 2006 um and it's the response like the War on Drugs is responsible for basically more than 90% of the Disappeared in Mexico in the last um like what is it like 20 years um so here it was basically um like the help needed was to analyze a specific Fields right uh where remains were found uh and we did that with um like liar or any other technology but also with uh models to predict like what what was the likelihood of find finding more remains or uh yeah like very very gruesome and it took me me basically uh like to go on the field to actually go

13:36 and search with families and with uh the authorities with the army with the National Guard uh in very dangerous places so the team with like the conasi team and the research centers teams like we will go with the families and with the authorities and actually go and find or look for uh people in like on the on ground um then the second one it was more about National Security we um wanted to provide or yeah to to to help uh build capacity on the graphical Information Systems uh do some causal analysis on crime and do fingerprinting for municipalities right so they could know like what what explained a crime or in in in the specific municipalities um and then something very awful happened which is the tragedy

14:39 of PL Lian um which um I'm sure most of you don't know but there's a very big problem in Mexico that the oil ducts are like punctured uh so people steal gasoline uh and then sell it like on the street right and to to do that they create a hole in the oil old duct and inundate uh pit and then they just with jars uh get the the gasoline right and what happened is that one of these pits uh took fire and it killed like 95 people right and the videos there were awful like you could see basically people borning alive and um yeah so we were at has to do like social profiles and help the federal police like with data processing and and stuff like that so it was a very um specific task force

15:41 but it took it also took the the ecosystem andal informatical methodology and then um the last one I participated with um was the health uh Annie which is like originally if was more about like um research on uh acute lymphoblastic leukemia uh due to widespread contaminants but also uh on rivers and pollution and trans transgenic uh corn and stuff like that right so it was more like friendly but then 2019 well the end of 2019 came and we knew right away that this was going to be pandemic um and I was asked to create a new task force for it um so that's the context of why we were doing this uh and now the topic of this presentation is um the co covid-19

16:45 ecosystem National information um so what we did very early 2020 um was a call for call for proposals like we knew that for instance the cat which is the the math oriented Research Center um had some knowledge on um how do you say that like um epidemiology like data data science for epidemiology right like they had modeled um Singa and other Chong guya uh and other diseases uh so they have um exper erience working with epidemiologic epidemiological data um so we basically asked them to to be ready and to create projects that could basically be plug and plate uh with the intention of

17:49 helping the government to make decisions on the on the pandemic um we received like 50 or so like very interesting Pro uh projects and we like we didn't say yes or no to anyone we just ask them like come to our meetings like a a standup meeting we did like um every day and we we'll see we we'll see from there right and at the end of um like the first month we only had like 15 20 uh very serious um projects uh but we basically had time uh to uh build a lot of relationships um and why did I had to build these relationships there's two answers right like one is the most obvious which is that um the pandemic was coming and the government needed information and insights right but also

18:50 the second one is because these researchers were doing this amazing work and dedicating most of their time uh Pro right so the promise I made to them is that um I will do whatever I had to do to bring their insights and their projects to the highest of levels right um so that's why I had to really really dedicate myself to building relationships with many uh government entities uh and to show them um you know like what we were working on I remember one day I think it was the second or third day um during like for Mexico's confinement so I was in El socalo right like which is the main plaza in Mexico City um and I met like very high officials in palas National which is the like where the president's

19:52 offic offices are are and then I cross a very empty socalo it was eerie to see like I there was no one there it was like ghost town and it like if you go now it's like I don't know 5,000 people right there right now but it was all all for me and I crossed diagonally to Mexico City's government offices right where I also show them the projection the projections and all the other projects that we we were working so um um we created a lot of projects like I said and gave uh tons of insights I'm going to show you like uh the most prominent ones because the way I measure prominence is uh true impact and I think these projects basically changed or had an impact on millions of lives um

20:52 because at the end of the day like the Insight we provided I was able to bring to the president basic basically um and to Mexico City's governor and you know like very high officials from the health Ministry Etc so yeah once we have um the relationships built which was not very easy because there's a lot of politics um implied but also a lot of competition uh many companies were already um know like in talks with the government officials um to provide them some kind of insights on how the pandemic was going to behave um but because this was from the the National Council for Science and Technology meaning that the top scientists in Mexico were behind this um it

21:53 was I'm not going to say easy but um it was you know like not as complicated uh to reach the like these ears and these politicians that really needed this data so after we built that we started implementing you know like in production all these projects and we wanted to ask this I would say three main questions um basically one of them like the first one is uh how many IC units will the government need uh by municipality in Mexico Mexico has almost 200 2,500 municipalities in 32 states um so we needed um um like a budget for each of them in terms of resources right so um we basically modeled um well I'm going

22:56 to explain how we answered this question after this but the second um question that we wanted to ask is um what what were the expectations right like how many people will die basically um which was a very gruesome question and how but we basically also needed answer this because it's it also takes resources right um and also the art T which is the the the transference factor for for any uh disease so for instance if I have uh covid uh how many people will I transmit it to right if it's one then the pandemic will be flat but it's more than one then it will grow and if in some cases um the RT in Mexico was like 15 17

24:02 right so for every person that was sick with covid they will uh spread it to 17 other people and so on and so forth right so it was fortunately it didn't um like right on time uh the first confinement was implemented so that basically stopped the yeah the the transfer of Co um but yeah together with rt and other factors um we did all these projections and these calculations and then the third the third one is like um we knew that I mean new in the sense in like we had a hunch that of course uh if you ask people to stay home the the spread of covid was going to to be St right but uh we actually wanted to measure it right and so we created a

25:02 project to do that I'm going to show you um which is the unified Mobility index by State this was very very interesting because it took data from uh Twitter Facebook and Google and then um based on GPS coordinates uh we did like uh Fair origin destination uh type of Matrix and Mees created a a baseline because fortunately we had data from years specifically for for Twitter so we were able to create a baseline right like so people in I don't know Mexico City moved this way uh before the pandemic and then we did we compare the actual movement uh during the pandemic with that Baseline and that was actually used by by the under Secretary of Health on a

26:05 daily basis he will say okay uh the state of Chihuahua uh has increased mobility and they should stay home uh Etc right like this was used to communicate with a lot of people um so yeah and I also I'm doing this talk because I think all these s that are amazing and they did the pro bono work um should get some recognition um so yeah I I I don't want to to miss the opportunity to thank Mario Oscar danela Eric and Sabino for this and they did they basically created this report um twice a week during more than two and a half years so it was a lot of work and and um they actually uh had a lot of impact but this was not um like the it was very well

27:06 communicated but I think uh one of the other projects that also provided uh very interesting questions and answers was um this vulnerability index by municipality right what we wanted to do is okay we know how many people are dying uh from covid or are getting sick uh in each municipality and we also had the census data right so we wanted to see which social demographical um variables explained mortality right so we did a partially Square uh analysis and we developed this um this index that basically was uh used to to focus or to create a priority list on which municipalities require ired more attention right um just a quick Insight on this we found that the education level uh had um

28:10 like a a very very big uh explanatory U coefficient uh for covid debts right and of course we we still need to or back then we still had to do more research on to why and to actually uh build uh the caal relationship um but to to me it was very very interesting uh and it made made sense right like um if the if the education level is very low then they like people don't have access that they should they also don't have access to health services uh they basically are abandoned by the government right because um when people don't have education level uh it it's not because they don't want to it's because they don't have access to services so to me that was the explanation um and this was done by Raul

29:10 which I who I also want to to thank um then the third one and I think uh one of the most uh impactful was the estimation for the epidemic curves um to determine risk right so basically this is the semaphor that the federal government used to say okay this is for this state they should be confined right or this state has improved so now people can go out and resume work um and what this chart shows like the one on the left is like we had the 32 states of Mexico uh on each row and then dates the um so you'll see for instance the the top one which is um oh it's not not only the 32

30:11 states it's also like regions uh so the the the second from the top is the national median so you'll see that on 2021 uh April 1st of 2021 until I think it says I canot see what it's basically one year and a half the the time series um how it behaved on on a national level but also on each state and when when things went from yellow to Orange that's when um the health Ministry um basically asked people to start a staying home right and then of course the red ones are when the cases were super super bad and when we saw jellow two green it was like a good sign right that and people started to go out but of course when people start to go

31:12 out then the cases start to increase so it became like one wave and two waves and three waves and we had many many different waves as we had as we are showing on the on the chart on the on the right which is um the time series for for Mexico City and how we calculated this uh like the the risk level um like I said this was used to determine if millions of people had or were were able to resume uh work or go outside um so in my opinion this was like super super impactful um this was done in semat which is the math oriented Research Center uh by technician um they are they were not like um you know like these top scientist that are in tat and who also work on other projects they were done as uh by Machine learning Engineers uh that work with

32:14 those scientists and they created this very very like very very fast so thank you Judith hin Domingo IET and ju um then the the last one that I want to show you this was done also like it was done in collaboration with semat uh the National University and um yeah and other I think it's uh UC Davis um so we like right in from the beginning of the pandemic we knew that we needed models right to basically say okay uh how Co is going to hit on a a municipal municipality and metropolitan area um level so they created this fantast like it like the prep print is amazing I love to read it it's super super interesting it's a basian model basically you put every PE every person uh in a box right

33:18 so it's like uh people that are are healthy people that have been exposed to covid people that got covid people that um got really sick with covid and needed a visit to the hospital than people that U required uh ICU or a respirator uh and then people who died right so based on that um and using this very sophisticated models we were able to inform the government with uh like at least three weeks of anticipation like where Co was going to hit right and I remember one meeting that I was with you know like the the top um government officials in Mexico in zo I was uh in my house I'm not going to say on pajamas but maybe I was uh and they uh like I was talking to

34:20 the top army guy and the National Security guy and all these like you know like top top officials and they were asking these questions right like where do we need uh icus where do we need to move doctors where do we need um you know like to put more resources and yeah um I think this basically helped Mexico to deal with the pandemic a little better um I'm not going to say that we did it in a good way because at the end of the day like more than for 100,000 people or 600,000 people according to some sources died in Mexico uh due to covid so of course like the ideal scenario will be for Ser Ser people to have died um but I think this model basically allowed the government to respond in a more efficient way and that

35:21 basically saved lives um and yeah this was also one um one of the the main like remember I told you that the promise I made to the researchers was that I was going to work as hard as hard as I could to bring this information to the highest levels um the moment I told them that my work was done was when the president was on a live stream and like he had printed this model like the we created a PDF that is actually open um uh this PDF and he printed it right and he was like on a live stream like in his desk like hey I have this information from the uh like the Mexico scientist right and we we are going to take made decisions based on this data and yeah so it was a Sunday um and in that that moment I told

36:21 them you know like I'm done and but I was not done uh so yeah um this predicted the wave uh the first wave of covid um which was very controversial of course right like uh it was communicated to the public because when you have this amazing Tool uh it's also very tempting to communicate it widely and to say hey the pandemic is going to end in MA May 26 I remember this date because it's going to that's that's going to be the peak right but of course people on May 22nd was saying like oh yeah it's going to end it's going to end and they started going out but we saw the first peak of deaths on May 26 and then a small decrease but then because people started going out they like a second wave which was more

37:24 much more stronger right so to me this model which was very powerful to make decisions like the ones I mentioned before like the UC UC seeu you yeah intensive care units and doctors and hospitals and whatever it was very powerful to do that but I don't think it it should have been communicated right um and it got covered all over the place and it was like super like the researchers which are these guys that are fantastic and I'm very proud to have worked with them well with all of them but in particular with this team um because yeah they have like yeah I'm I'm very excited to to talk about this um they they got bombarded with media requests so now I had to deal with the media too because basically we asked them not to talk to the media right like it it's not convenient for the

38:25 researchers to directly talk to the media um basically during an emergency right so I had to deal with that and control the situation um but yeah that's I think it came with the territory um so yeah just for the record I want to thank Marcos Antonio Andres and Maria Louisa uh for this fantastic and very impactful and very be beautiful like I I recommend you to go to the website which is um coronavirus do on AIT MX and there you'll find the like all these research uh products and PDFs and data like it was a a very beautiful open data portal for this nasty so just because I I think I only have a minute I want to talk a little bit about the impact well I I've already told you a little bit right but um my God where is it

39:28 oh yeah this um like I said critical decision decisions were made using the the insites that we provided um one of the like the secondary effects of This research was that um people were able to basically had lever or or lever on to measure like Health versus economic uh or health versus economy right because confinement um even though was the ultimate solution it's not economically sustainable right like people especially in Mexico and other countries uh live day today right so if they cannot work they cannot eat right so we had or the the government had to basically balance between that um and yeah like I mentioned mexic president presented one

40:28 of the of our reports during a live stream and quoted it multiple times during his Mana which is a daily um press conference that he does um then on a daily basis we worked with the health Ministry um to like be they had they also had a daily press conference uh and before they went online we had a meeting with them to see like the status quo and yeah basically it was like a standup meeting but with the under Secretary of Health which is Lopez gatel who I also want to thank for the opportunity and the trust that he put in the team um and we did that for more than 18 months so it took a lot of resources from our teams um and we also created very many many adhoc analysis that were not actually published in because there were like very specific um

41:30 and yeah mediatic the data was covered by most of Mexico's um newspapers but also the New York Times Did piece on it um and many other media Outlets so I think it was a very impactful uh ecosystem National informatic and with that oh this was supposed to say thank you but thank you uh thank you so much uh we have time for a couple of Q&A who wants to make one question please raise your hand and I'll let you lend you the microphone hello rul hello when I hello um I'm so curious about

42:32 your comments about how to best communicate model predictions to the public um and I think it's going to come up more and more I would love to hear what you took away from that about how to think about that version of Public Communication of Science and risk when you're doing that kind of predictive modeling so yeah I I mean the main learnings I I got from this is that um well collaboration with the is key right and um I think we need to actually make a really big effort to build Bridges and to uh not only do the work but also communicate it and to bring it to the right ears um and then also for instance uh if it's if it's correct to do like we can also communicate to the public um which is

43:32 not very easy right like uh if I go back this where's where's to the chart on the right um normal people think that the red line is the actual Peak right but it's the median and it's a probabilistic um Range so we have the P10 and P90 right and that's not very easy to understand or to translate into public policy right so I think the key um or the role or the learning I I I take from this and why I'm so proud of this work because I became the translator right like I I was able to understand these guys but also to collaborate with people that had no um no background around on statistics and you know like so yeah I think yeah the the the main learning

44:35 here is uh Building Bridges and VI liason Leon you say so we have time for more questions uh thanks for sharing all of this work with us um I was wondering if you had time to kind of go back and evaluate some of the models that you used to to see if they were um you know close to what what happened in reality um unfortunately unfortunately we didn't uh did a postmortem right I think it's do like it should be done right to see um um to see with the current data like how will the models behave um we didn't have

45:36 time to do it uh so but that doesn't mean that the the models were not like tested and actually all the uh like these models need to pass all the hypothesis testing like in a stat statistical way uh to be statistical significant right so they were U tested and soundproof just to just to say um but they were not like we didn't do a retrospective thank youa PA for for the

46:52 consolid well I'm I'm going to translate the question um like after is what's the impact in kit right like if the area if the area we created uh and Seline helped create create that like we work together uh for two years one um if it's if it it's going to stay right like and the answer is yes like uh we fought very very hard to basically change the law so they will allow us to create this area formally so now they have like uh resources they have um plazas which basically mean like uh positions for directors and general directors and coordinators and it's a I think it's a mediumsized um mediumsized organization that took almost three years of my life to create um but yeah

47:53 it's going to stay I suppose um I don't know if they like after I left I left in by the mid 2021 um I don't know if they they released a couple of ecosystem naal informaticos on other topics but I don't think they did attack like the Urgent issues that or the pressing pressing issues that we did it's different but respond to when [Music]

49:25 ACC well sorry I'm I'm not going to translate it but basically it's about the Disappeared people um and how we like what's the results of that right and um okay um

50:55 what's the impact I think the the outcome of this was a DAT database it's I don't think it was publicly or or it is publicly accessible it's basically intended for use for researchers only um and but there's a couple of papers published on on this topic uh by the like the person from the commiss naal and uh people that were on my team right so they published that and I think the the database and the system really really works and has helped find or give answers to these these questions thank you so much paa much gracias uh please give a warm Applause to paa thank you thank you thank you um it's now time for the launch uh

51:58 and we are back in 1 hour I think so here right thank you so much