With two patents and an IEEE publication, Meta’s Isha Oke is on a mission to democratise data science
Our Women in Tech series features Isha Oke, Data Science Lead at Meta. During her stint in Microsoft, she received two patents for inventing new ways of targeting devices and transforming data into features for ML.
Growing up in an academically driven family, Isha Oke was inspired by her grandfather, who was a big proponent of Applied Mathematics. Years later, the Data Science Lead at Meta would realise how this innate love for Mathematics would mean much more than doing well on tests or at mathematical competitions and instill a sense of purpose.
“Through my grandfather, I understood Mathematics is useful in every walk of life, no matter what industry you choose to be a part of. Though he introduced various application-oriented methods of studying the subject, apart for being good at it and enjoying it, I did not feel I was doing something meaningful,” Oke tells HerStory on a call from Seattle, USA, where she is based.
“Now, in my role at Meta, I realise I am using the skillsets based on applied mathematics to make an impact on millions of businesses at once,” she adds.
At 28, Oke has two patents to her credit from her stint at Microsoft and has also contributed to an IEEE publication on developing an adaptive learning system personalised to the learning style of a student using machine learning. She worked on this while volunteering at Pune Learns, a community focused on teaching spoken English to underprivileged children to make them more employable.
Oke chose computer science and information technology for her bachelor’s in engineering at College of Engineering, Pune.
“It was another kind of applied methodology of using mathematics and logic, and that intrigued me,” she says.
The turning point
However, she says she felt lost initially despite studying cloud computing, networking, operating systems, and almost everything under the IT umbrella. The turning point came during the final year, when she opted for Business Analytics as an elective.
It changed both her perspective and the path ahead.
“Our curriculum was designed in a way that helped us understand not just what tools are used, but analyse data and apply strategic learnings and recommendations to make operations of a company more efficient,” she says.
When Oke graduated in 2016, there were very few jobs in Data Science in India. According to her, at the time, data science just meant machine learning. She was specifically interested in product data science and decided to move to the US to pursue a master’s in information systems management at Carnegie Mellon University.
Two patents to her credit
In 2018, she joined Microsoft as a data scientist and soon became involved in every role “that could fall under the umbrella of data science.”
This included data engineering--figuring out how to collect data from different logs, sanitising the data, putting it into pipelines, and building the right tables. With machine learning, she learned where to use that data and develop models to solve business problems. As part of product data science, she was making strategic and statistical recommendations on what product to launch or not. It helped her choose what she wanted to do next.
During her stint in Microsoft, Oke got two patents approved.
She explains, “As part of my team at Microsoft, I was on the Windows and Azure security team, and we developed, as my first patent, a methodology to automatically apply security policies to different Windows devices, using machine learning.”
The second patent, an idea Oke conceptualised from start to finish, was to identify a big problem that existed in the organisation.
Oke noted that every week, Windows users submitted feedback through a tool called Windows feedback hub that existed on every user’s device. It could be something actionable like “I had this app crash today. I don’t know what to do.” Or something non-actionable like how I can get free Microsoft Outlook without paying for a subscription.
“We realised we were spending almost four hours a week, triaging this feedback, and marking them as actionable versus non-actionable. And 70% of the feedback was non-actionable.
I realised there is a lot of training data and we have been annotating this data manually for years. So, this is a very handy machine learning problem for us to tackle,” she says.
She created a tool that automatically triaged this feedback within Microsoft, and marked it as actionable versus nonactionable.
This tool has now been deployed within Microsoft's internal feedback triage process and is available for every team to use, saving 10% of developer productivity or four hours per week for each engineer that triages feedback.
Oke believes her stint at Microsoft helped in her deeper understanding of data science.
“In my first few months at Microsoft, I thought it was all about creating these fancy machine learning pipelines using the most sophisticated machine learning algorithm and applying different analytical tools to your data sets.”
She understood that it doesn't matter what tool you use and what level of sophistication you have in your machine learning models, as long as you're solving a problem.
Building solutions for SMBs
In February 2021, Oke joined Meta as a data scientist and was promoted to Data Science Lead within a year. She started off on the advertiser products team that built products and offered solutions for advertisers, specifically small and medium businesses (SMBs) that don’t have much marketing expertise.
“Most of the businesses on Meta’s ad platforms are SMBs, usually one-man shows, whose founders don’t have the time to take a marketing course and find the most optimal solution for advertising. Especially during the pandemic, when these businesses were hit the hardest, we wanted to make sure that their advertising efforts were optimal and fruitful,” she explains.
She cites an example where an advertiser was looking to deliver his ad through Instagram stories or reels. This was holding them back from cheaper opportunities that existed within Meta’s ecosystem. The team guided them on choosing more locations to deliver their ad for better reach and impressions.
Oke’s responsibilities in Meta include identifying future investments in reels ranking/ monetisation through data analysis, evaluating tradeoffs between different launches, metrics, and strategic decisions, defining the right metrics and rollout plans to measure success of different products through complex A/B testing setups, creating a culture of data-driven leadership through keynotes and mentorship programs and also mentoring new hires and colleagues transitioning into Data Science at Meta.
She also speaks about being part of a project that was credited in an Institution of Electrical and Electronics Engineers (IEEE) publication. This happened while she was volunteering at Pune Learns, a non-profit that upskills underprivileged children in English communication.
“We found a lot of kids were dropping off from school after their 10th and 12th standards. We realised that if we upskilled them in English proficiency and English communication, then they would have employable skills and access to jobs. It attracted a good turnout of students,” she says.
But, the problem, she realised, was at the grassroots level, as every student has a different way of learning something.
Some of them prefer a visual medium, and some of them prefer verbal or textual medium. And there isn’t just one way of teaching a student, but it’s more fruitful if you cater your teaching style to what the student is used to receiving, or what the student is more likely to respond to.
“It is a difficult problem to tackle at scale. One teacher cannot modify their teaching style purely based on the individual learning style of every single student in the class. We developed an educational system, think of it as a web app, where there's training material, but based on the inclination of the student about whether they are more driven towards visual mediums, or towards text. Using this, we were able to deliver the content to them in a more tailored way,” she adds.
As a woman in technology, Oke believes she’s been lucky to have female mentors in different times of her career and learn from their leadership styles.
“One of the female leaders taught me how emotional vulnerability and empathy can help build stronger relationships with mentees. By emotional vulnerability, it means it is okay to be extremely transparent and honest about your strengths and your weaknesses. And it’s also okay as a leader to acknowledge where you're lacking, and where you can do better,” she says.
(The story has been updated to change Isha's age from 29 to 28 years.)
Edited by Megha Reddy