Omar Florez is our Technical Panels & Workshops Deputy Chair but also an invited speaker at Tapia 2021. Omar is currently a Machine Learning Researcher at Twitter Cortex. Omar is also part of Latinx in AI and we spoke to Omar about his involvement with Tapia conference.
Tell me about where you grew up and what work your parents did.
I grew up in Peru, I studied computer science there and did a lot of research when I was an undergrad as I was very curious about concepts like entropy, pattern recognition, and gradients. Doing research as an undergrad was fun and I learned a lot. Later I figured out that doing research was good to get into graduate school. I applied to schools in Chile, Brazil, and US for computer science programs. I did not know at that time but without publications would have been hard to be accepted.
Went to Utah state university to finish his PhD. Peru to Utah and doing publications.
My parents never went to University, they are merchants in Peru. When I was young, they told me to not focus on money but on knowledge. In Peru we often have earthquakes, so people can easily lose their assets due to natural disasters. When they encouraged my sister and I to study and not helping them, they just wanted us to learn as much as we could in school hoping that will give us a different life.
How did you get started with computer science?
Got my PhD in Computer Science. Went to Tapia for the first time in 2009 – met Patti Ordonez. We were both attending the Doctoral Consortium. Knew each other there. Became friends after that. Next Tapia won the best poster award which allowed me to have more opportunities while I was looking for a job. Got an offer as a research scientist at Intel Labs in California and was there for 5 years. Then I went to Capitol One as a Senior Research Manager. Recently, I joined Twitter as a Research Scientist.
I never did a PhD having about a particular job in mind. I did it to get deeper knowledge about areas I was curious about – AI and Machine Learning. That was the motivation for doing the PhD. Having deeper knowledge and expertise to understand how we can use Math and computers to predict things. Later I realized that this mindset could help you in your daily job as well.
What are you working on now?
In general, I have always worked in Machine Learning. PhD was in predicting abnormal configurations of vehicle and people interactions using surveillance cameras to prevent accidents on traffic roads. IBM Research gave us an Innovation award on that research. Right now, I study how we can use unlabelled collections of data to understand how people communicate in social media. We need lots of data to build these models. However, people communicate differently in social media – emojis, memes, entities, news, I study how to use all this information to understand and generate content hoping computers will help to decode human language.
How did you become involved in the Tapia Conference?
Doctoral Consortium and then Poster. Patti is Still a friend. She became a role model for him – now colleagues. We often try to work together to help people. Even have spoken to her students in PR via videoconference.
Tapia has left a mark in me since he was a student. First time I have seen such a large group of Latinos and Latinas in a CS conference. Before that I used to attend pattern recognition and machine learning conferences and met very few Latinos and Latinas.
Patti and I have been inspired by Tapia creating similar efforts in own communities. For example, I am part of Latinx IN AI which help Latinos and Latinas from all over the world to attend top Machine Learning conferences such as ICML, NeurIPS, and ICLR.
Many people he met at Tapia – have also helped our community in many ways. Tapia Conference is a good place to find people that are oriented to help others.
We are celebrating 20 years of Tapia Celebrating Diversity in Computing this year. What are you looking forward to this year?
First time he attended Tapia I got a message – together we are stronger. Learned this and I use It in my own life all the time. Achieve this year at Tapia – we Latinos and Latinas – can be closer despite of current restrictions. This is just a reminder that we can help without being physically in the same place. I am in CA and Patti is in PR, Texas, Colombia. If we want to help students – we can use the internet to create opportunities, the Internet can help us to create a vibrant and global community.
What are your hopes for the next 20 years of Diversity in Computer Science?
When we talk about diversity, we should consider the full range of its meaning. Diversity in the access to opportunities. Diversity to learn. and engage despite of language. I envision the next 20 years of Diversity in Computer Science as a moment in which everyone will have the same opportunity to learn. For example, a homeless on the streets of SF or a housewife in a Latin America will be able to learn math, English, python – independent from finance – if they wish to study more. If we have way to connect everyone to internet and give them free access to computing resources, there will not be excuses to learn to code and build machine learning models. My dream is that if someone works hard and has internet connection and time, there should be a path to prosperity without even going to university and independent of where you were born.
Right now, we are somewhat constrained by language – maybe someone in Peru might want to access latest research and papers but it is only in English. Machine Learning can help us a lot, in few years we will have computers that can do fully reliable automatic translations allowing everyone to learn, contribute, and engage to the ML community despite of language.
An advice for students at Tapia.
Learn as much as you can and let curiosity guide you about what to read next. While learning a lot, try to teach those that lie behind or are still beginning their trip. This is important because we need of everyone to solve some of the difficult machine learning/AI problems we still need to solve in the next years, for example neural reasoning, causal models, zero shot learning, etc. The more people we have in AI, the more creative solutions we will have to figure this out.