Today on Tech Sisters Stories we’re excited to have Zahraa Al Sahili.
Zahraa is an incoming PhD Student at Queen Mary University of London and Google DeepMind PhD Scholar. Zahraa is also an AI Session Lead a Udacity.
Listen to Zahraa’s Story
Key Lessons from this Episode
- Zahraa’s experience studying in Beirut during the height of the economic crisis (2:32)
- Zahraa’s favourite AI project (7:30)
- The complexity of removing implicit bias from AI (10:00)
Transcript
Zahraa Al Sahili
[00:00:00] Grace Witter: As Salaam-Alaikum, you’re listening to tech sisters stories. tech sisters is a community that connects you with other sisters who share your story experiences and goals. So you no longer have to feel like the only one like you on your team. My name is grace and I get to interview the amazing women in our community, share their stories and the lessons they learned.
[00:00:22] Grace: as salaamu alikum today on tech sisters Stories, we are super excited to have zahraa al sahili. Zahraa is a current PhD student at Queen Mary University in London, and she’s also Google Deep Mind PhD scholar.
Zahraa is also an AI session lead at Udacity, and I’m so happy to have you on today
[00:00:42] Zahraa: Thank you so much, grace, for having me here. It’s my pleasure.
[00:00:45] Grace: alhamdulillah. So how did you first get into Tech, Zahraa?
[00:00:48] Zahraa: So as a weird starting first, I did my bachelor in electrical engineering after some motivation to math and physics. And after graduation it was a tough period. Lebanon was passing with no hiring and no jobs. So how everything started is honesty through coursera.
Yeah, so my bachelor was focusing mostly on power systems and control systems and radio frequency. And then I started building up my machine learning skills through coursea honestly through Stanford’s machine learning course, and then the machine learning specialisation.
And after that, I joined American University of Beirut as master’s student, but now focusing on artificial intelligence.
[00:01:29] Grace: MashAllah, so you are very academic, very driven, mashAllah, and I love that we’re using Coursera because that just makes these certifications accessible to everybody. And we’ve seen a lot of tech sister members use coursera in, in place a university degree. So I’m really happy to see you’re able to combine those.
[00:01:47] Zahraa: My pleasure to start my journey there. So coursera offered us very updated curriculum, just like Stanford machine learning courses. And also for free. So during that time we had a financial crisis in my country. So we were offered to apply for financial aid and took all the courses for free.
And honestly, I built all the knowledge I really need to go into machine learning research through these online courses. And since my day one at the master’s program, I start working on machine learning research. And the benefits were the masters are mostly just to enhance my research skills, but the knowledge at coursera was really very sufficient to build a very strong and powerful machine learning engineer.
[00:02:32] Grace: mashAllah. That’s really great. I’m really happy to hear that. Let’s talk a little bit about. The crisis that Lebanon has been going through. So of course it’s very, very difficult. We’ve talked to a couple of different people focused in ai, in Beirut. What do you think is something that needs to happen to help more people in Lebanon be able to get jobs and to have sufficiency, especially in the tech sector?
[00:02:55] Zahraa: So I think first supporting students and enhancing their technical skills. So I think one of the advantages of the Covid Pandemic is opening the opportunity for remote working. So currently, although we are having a problem in working in the remote or tech sector as Lebanese with Lebanese companies, we have the opportunity to work remotely abroad.
So I think the most important thing is to build efficient skills, and as I said, so either through coursera or with support of the government. I know other countries are supporting the students, for example, ATAC through government funded program. So I think having these steps since in tech, there is the opportunity of working remotely.
The second problem is I think the electricity, which is really. Very affecting us as computer scientists or anyone that really needs electricity for his job. I think also giving some support to, to students or to graduates and solar panels or power banks for those who are not really able to afford this.
The internet also is another problem, but I think with sufficient skills and just resolving the issue of electricity, or at least some companies understanding this problem, like at this from just say 2:00 PM to 6:00 PM I really can’t work. I have cut of electricity. So I think this also can help.
[00:04:22] Grace: Yeah, that’s it. mashAllah, I think the infrastructure especially was what I was, was thinking of. That’s one of the things that we heard from the other people. We interviewed from Beirut and we also were talking to Manara who does programs across the MENA region and infrastructure. Something that is just very difficult to deal with.
So having enough electricity. Supporting students with getting the laptops and the hardware that they need to do their job because it’s it’s a lot of money. subhanAllah, yeah,
[00:04:51] Zahraa: And we were having it even at, let’s say my university, we didn’t, we sometimes had some even computational. Problem. Let’s say we as machine learning engineer need the GPU to train the models we have. And as you can see, if you just check the projects of machine learning researchers at the Lebanese universities, they are restricted to some type of topics that can really work without GPU.
Since we didn’t have this, not since, we don’t like to go for. Projects that are interesting or difficult, but since we didn’t have enough computational resources really to go throughout these projects, so , this was really another challenge and the university just respond that we really wanting into a crisis and GPUs are very expensive and we can’t really find you with the GPUs to go for your research.
[00:05:43] Grace: So now doing your PhD at Queen Marys, are you still, are you doing it remotely from Lebanon or have you relocated to
London?
[00:05:50] Zahraa: no, I just relocated
[00:05:51] Grace: to London
[00:05:51] Zahraa: like, you know, after all these problems, I think doing a PhD remotely is
[00:05:56] Grace: It would be very hard.
[00:05:57] Zahraa: with a problem with electricity Yeah.
[00:06:02] Grace: mashAllah. So how is the, the comparison between the two? Is it just a huge difference between Queen Mary’s and Beirut?
[00:06:10] Zahraa: I think the support, but I was just like about my previous university, which I really living here again, was the lack of discrimination.
So, and the one, mostly we have discrimination against the. Female wearing hijab, or especially in the recruitment sector. But at this online university, we were supported to the extreme.
There was no discrimination at all. And the same here. So even let’s say in the PhD, everyone is very nice and everyone is very supportive. I just had the more, the advantage of having powerful GPUs to train everything I do more
[00:06:46] Grace: Did you Oh, that’s wonderful, alhamdulillah. And what do you hope to do next after your PhD? Zahraa?
[00:06:56] Zahraa: So after my PhD I’m planning to become a machine learning research scientist and with a motivation to AI for good projects. So actually the most. Beautiful thing about AI that’s really motivating me to do my research is the good motivation. So working on AI for good projects, which I worked on a lot in my masters and now also in my PhD.
And I’m planning also after becoming a research scientist to help improve the artificial intelligence field in the direction helping in the good of humanity.
[00:07:29] Grace: Oh, mashAllah may Allah make that happen for you and enable you to do all those wonderful things inshAllah.
[00:07:35] Zahraa: Thank you so much.
[00:07:37] Grace: What was a machine learning project that you were really excited about? Something that was a really nice puzzle that you had to think a lot about it. It was really challenging, but you’re very happy with how it turned.
[00:07:47] Zahraa: So I’m talking about a project that was not that difficult, but I’m really very proud of. Helping others a lot compared to other very sophisticated projects that I contributed to is the impact of my, one of the projects related to agriculture. So I really, especially during the crisis that we are having, you feel with the farmers, you feel with the poor people, you feel with everything.
And that project was basically to help farmers detect, plan the diseases pests and weeds for free and The thing that really excited us is that we tried, I put mainly all my effort to make this model as much as accurate as possible, just in the sake to really help farmers when deploying the project and saving their crops without, since they really not able to pay for agriculture engineers or expert to save the field.
So that’s really one of the projects that really means a lot to.
[00:08:46] Grace: And like you were saying before, this is something that has a direct positive impact on people. So yeah, I can see that mashAllah. And were you successful in, in your model, was it able to predict what you wanted?
[00:08:57] Zahraa: Yes. So it was very accurate. We also published it in the frontiers and plant science, and now we are in the process of building the mobile app of the machine learning models, so to deploy them for farmers. And we tried our best to make it as much global as possible, so we collected all the.
Public data sets over the whole world so that it really helps farmers from different countries. And this also have an impact on making the models more accurate and more generalizable.
[00:09:26] Grace: MashAllah Zahraa. That must be very satisfying to see your project. Into something or is close to being a mobile app and then having a, , big impact.
[00:09:36] Zahraa: Yes, hopefully. And it’s coming to save a lot of small farmers,
[00:09:40] Grace: so one criticism that often comes up with AI is the implicit bias that can happen when models are being built which often happens when it’s being built with non-diverse teams. So the AI machine learning researchers are not coming from diverse backgrounds, and therefore the models have a bias that’s implicitly built into it.
Do you think that happens? How do you think that can be remedied? Have you seen something like that in your studies?
[00:10:09] Zahraa: Um, Yes, and we as minority, these projects means a lot to us. So before I worked on bias mitigation and hate speech detection systems and for different, and entities , whether related to religion or gender or origin of the of the tweet poster. Since, you know, let’s say for example, I’m just saying a tweet and usually the moderator reported that I hate Syrians or Syrians are bad.
And then the special identity past the tweet act. I am Syrian. And then the tweet is reported as hate speech by mistake. So currently till now, It’s very difficult to remove this bias at all since, you know, the bias, as you said, is from different origins. So first, we have the bias in the data. The data at substance.
We are biased as humans. The bias in the data collection, the bias inside the model, the bias for us as AI engineers but mitigation, of the bias is, is currently that the possible solution. We are having some algorithms. Try to remove or reduce this bias, especially let say for manually adding these possible tweets that, that say are to be biased or let’s say transfer learning, multitask learning.
I’m very excited to this direction. Or hopefully we can reach an algorithm that really is able to remove bias and not only trying to reduce this bias, and let’s say the bias was reduced by 20% or 30%. Hopefully at one extent we can reach a limit where really this bias is removed at all.
[00:11:47] Grace: inshAllah. That would be wonderful. And it’s, it’s interesting hearing the different sources of bias and how deep it can go. I think a lot of people think that it’s just something that you can clean up really easily or it’s something where you just have to pick the right data sets, but it is quite complicated.
[00:12:05] Zahraa: Yeah, it’s very complicated. It’s very hard to remove, but many researchers and thanks God are dedicating their time to resolve this issue, and hopefully in the next upcoming years we can reach a place where really this bias is removed.
[00:12:22] Grace: And Zahraa, what’s been your experience so far working as a Muslim woman in tech in this field?
[00:12:28] Zahraa: So mostly I worked in the research domain as in field of computer vision, as I said in agriculture and AI for medical imaging. Also an NLP on mitigating bias and hate speech systems. And finally, on pandemic forecasting and graph neural networks for pandemic forecasting and parallel. I also worked as a teaching assistant and instructor, so in AI summer schools, and also now at Udacity, I am supporting Amazon Scholars and AI journey.
So in the. Programming with Python Scholarship Program and expanding. So democratizing the education again. So as I received the education from coursera for free most now supporting Amazon scholars to learn AI and pave their journey to help humanity.
[00:13:16] Grace: mashAllah, that’s wonderful. I’m really happy that you’re able to pay it back like that. What do you, what advice do you have for any students who are coming into your field?
[00:13:24] Zahraa: First thing, never give up. Always learn. So this field is moving very fast. So always update your skills, let’s say every one or two weeks and never really let anything stop you from learning or acquiring skills. So we have, in coursera, we are having edX, we are having all universities are publishing their courses on YouTube for free.
So if you don’t have enough money, let’s say to join any graduate program or to this very expensive starting machine learning program. Just learn online and you’re going to acquire the skills that lets you really pave the way, and then don’t let anything that you give up or stop you from learning.
[00:14:08] Grace: Zahraa, what is something in your career or in general that you’re most proud of? So you’ve mentioned your project before for small scale farmers. Is there something that’s like really close, dear to your heart?
[00:14:19] Zahraa: Yes, so mostly that project is the most one closer to my heart, also the AI mitigating bias. And finally, the Covid Pandemic projects. I work on multiple covid pandemic forecasting projects just to also help. People. So introducing the number of cases of Covid. So this forecasting algorithms help governments decide whether to close or open and really affected the spread of the disease and the lives.
You know, basically before the vaccination really spread of the disease was also affecting types.
[00:14:52] Grace: mashAllah, that’s really important work. . What’s something in your career journey that you regret or you might wish that you have done differently?
[00:14:59] Zahraa: Basic, honestly, I’m thankful for everything even. Things that were not that good in my journey or the the bad accidents or the, let’s say, bias accidents that I faced before. So this made me more stronger. Also supported me to decide on moving outside and starting my PhD. So they were really strengthening my skills.
So I think without these bad things, I never imagined working outside or pursuing my PhD abroad. Also they made me a stronger researcher. When you’re having these bad incidents or accidents, this enhance your skills and help you become stronger.
[00:15:40] Grace: And Zahraa. What’s something or someone that you’re most grateful for?
[00:15:44] Zahraa: I must grateful for, for Antoinette for sharing his courses, public . And also I am very thankful for Professor Haha from American University of Beirut to help me in this way. So when I was a power engineer, I just talked to him and he guided me to course and how to start my machine learning journey, and then how to enhance my research class.
And finally in supporting me in the PhD application as a recommender so I’m also very grateful for him.
[00:16:14] Grace: What a wonderful mentor. That’s
Alhamdulillah, is there anything else that you’d like to add?
[00:16:21] Zahraa: For anyone who is interested to join the machine learning research community, never let anything stop you from achieving your journey with all these rumers that you need the PhD degree or it’s very tough to enter to the field without the graduate studies.
I know a lot who really ended up doing very exciting work with only online courses. So don’t let anything stop you from entering the field, whether your financial situation uh, whatever, anything else like in Lebanon, we had no electricity, we had problems in the internet, but you see, we just having this thing to join and and enter the field and hopefully with these challenges and struggles, you’re going to excel and become stronger researchers, and then more opportunities are going to open for you or maybe scholarships or jobs upwards if the vacancies available at your country doesn’t satisfy your ambition.
[00:17:19] Grace: Oh, perfect. That’s what a great sentiment. Thank you so much, Zahraa, for, for coming and for sharing your journey.
[00:17:27] Zahraa: Thank you so much, grace, and I’m also here always to help anyone who’s interested to join the field,
[00:17:33] Grace: Yes, definitely. So you are in our Slack channel, so anyone who is interested in in machine learning anything like this, they can reach you and talk to you through
Slack
[00:17:41] Zahraa: always help here to help.
[00:17:43] Grace Witter: And as always, thank you so much for taking the time to listen today. If you liked it and you like what we’re doing at Tech Sisters consider following us, leaving a review, sharing this episode with any friends or even supporting us on Patrion. All of those really help us a lot. This is a completely non-profit organization. We’re just doing this for.
Sadaqua , so anything that helps more Muslim women find us and discover us and hear the stories is immensely helpful. And if you are a Muslim woman in tech, please go ahead and check out our community. It is completely free and fun and very supportive. You can join by going to our website tech-sisters.com and filling out the membership form, and you will get a link right away into our slack. So it’s really, really easy.
And that is all for me. And I’ll see you next week. As Salaam alaikum.
