What technical or personal skills help distinguish yourself from your peers within such a competitive industry?
Kesher: To be fair, I am not your quintessential “unicorn” Data Scientist, which I think is more a myth than a reality. I am someone who is interested in understanding how the world works through the lens of data, and present my findings in a format that everyone understands. The world of analytics always had very smart people and more recently has seen an influx of highly talented folks. This has led to a hyper-competitive space where it may be difficult to stand out and differentiate yourself. In order to distinguish myself, I try to keep abreast with the latest and the greatest technological advancements in the field of data sciences, as well as keep improving softer skills such as storytelling, data presentation and communication.
I also feel that my education leading up to graduating with an engineering degree provided me with a great foundation in linear algebra, logical reasoning and statistics, and my professional background in software engineering has helped me pick up any programing language almost instantaneously. This combined with the soft skills I developed during my career, as well as while getting an MBA degree, has helped me become a data-driven storyteller. I am a big believer in the notion, “learn something new every day,” an incremental development in terms of learning a new language, or a machine learning technique, or a new visualization format to tell a compelling story.
The term ‘innovation’ has been increasingly used by software and tech companies to describe their ventures and aspirations. What does innovation mean to you and how do you apply this to your work, specifically with data science?
Kesher: To me, the term ‘innovation’ means a laser focus and an unwavering commitment to what’s good for the people that we are building the product for. For example, analytics is crucial to the success of every product, but smart data use opens new avenues for success and improving the customer experience through personalization and relevance. Another aspect of innovation is the way data science has led to product improvements being measured, quantified and tracked – you cannot improve what you cannot measure. With data science techniques, we are now able to measure and improve product innovation at a tremendous pace.
What was the experience of speaking at Innovation Enterprise’s Big Data Conferences like for you? What were some of the main points you wanted audiences to take away from your presentation?
Kesher: Innovation Enterprise’s Big Data Conference is at the forefront of bringing together an exciting agenda packed full of presentations, workshops, training sessions & panel discussions. It presented an amazing experience to contribute and share data strategies from industry forerunners, meet with potential recruits as well as network with leaders cross-industry who are connected by a common desire to use data to enhance people experience. The theme of my presentation was ”Making Sense of Unstructured Data.”
The key points that I wanted audiences to take away were how to understand unstructured data and realize that the next wave of innovation will come from uncovering insights hidden in unstructured data, which includes images, free-form texts, and raw sensor data. I wanted to give the audiences guidelines to make sense of unstructured data that have worked for me. Also, how deep learning is revolutionizing the understanding of unstructured data
What does data science mean in regard to artificial intelligence and machine learning?
Kesher: Data science has become a broad term, and encompasses data scientists that are covering a wide range of roles like machine learning scientist, data engineer, data analyst, data infra engineers and business analytics professionals. These roles can range from being more technical to more product and business-oriented. For example, the more technical roles in machine learning and data engineering, including infra, require a sound software engineering background plus an understanding of advanced math and statistical concepts, whereas the less technical roles need good domain expertise and skilled data manipulation knowledge, like SQL for example. What you end up pursuing should largely be determined by your interests and skills, and there is ample opportunity to produce product impact in any of these roles. At the end of the day, it’s the actionable insights that matter in moving the product forward and solving real-world problems for the people using the product.
How were you able to utilize data and analytics to increase efficiency and/or the sales of both Fire TV and the Amazon Echo while working for Amazon?
Kesher: Utilizing data and analytics to inform business decisions was a core aspect of my role at Amazon. As far as using data and analytics techniques, we utilized them for sales planning through creating forecasts of sales which not only informed our pricing strategy and bottom line profitability but also helped with building plans and inventory accounting. They also helped us understand our customer experience. We created usage graphs from clickstream and derived insights therefrom to build better products in addition to lifetime value modeling in which we used regression models to develop the downstream impact of device ownership.
Given that many of the specifics and code behind artificial intelligence and big data is specialized, how would you go about democratizing them?
Kesher: The surprising part about data science knowledge including availability and dissemination is that it’s more democratized. There is a very active research community and the academia has been at the forefront of creating cutting-edge algorithms for over many decades now. However, it should be noted that it’s only now, with cheap computers available through cloud platforms, that we see a breakneck pace of implementation.
As for me, I am doing my part by sharing non-confidential knowledge and the best practices at prominent AI and Data Science conferences where I have been invited to present and talk, reviewing scientific presentations and contributing to editorial content reviews for prominent data science journals. It has been very fulfilling to me because it gave me opportunities to discuss my experiences as well as give back, learn from peers and explore new ideas and directions for future work.
Are there any new applications on uses for AI that you see being implemented in the near future?
Kesher: This is a tough one! While there are many fields that stand to benefit from the proliferation big data and artificial intelligence, the one that I am most passionate about is alleviating the water crisis now and in the future. A little bit of personal history, my senior year ”major project” was using neural networks to optimize hydro-dam inflows and outflows. Water has become our most critical and contested resource and I would absolutely love to see technology help us measure, plan, and more effectively allocate precious water resources where they are needed most, such as for fresh drinking water and food production. Deep learning technology that has helped increase the efficiency of oil explorations can help analyze satellite imagery and use new-gen technologies to monitor surface water levels and fluctuations around the world to help inform the strategy on water utilization and conservation.
Over the course of your time at Amazon as well as being Facebook’s Head of Marketplace Experience Data Science and Head of Platform Data Science, what would you consider to be the proudest moments of your career?
Kesher: I am the sort of person who likes to live with many small wins and as such, I don’t think there is one moment in my professional career that qualifies as the proudest. However, I must say that every time I see friends and family use the products I helped to build, from Echo and Fire TV at Amazon to Facebook Marketplace, I do a quiet “hurrah” dance in my mind. To me “real usage of the product” is the best form of flattery and accomplishment, and I hope to keep building awesome products. I would like to leave the readers with a shameless plug that if you are looking to create history and build awesome products by leveraging data science, please get in touch with me. I am always hiring for Analytics at Facebook.
What were the key takeaways from your article in the ORMS-Today by INFORMS and why do you think the topic you covered was significant and/or interesting?
Kesher: Authoring a featured article in the December 2017 edition of the ORMS-Today publication by INFORMS was a great honor. As you may know, INFORMS is the leading international association for professionals in operations research and analytics, with over 12,500 members worldwide. They promote the best practices and advances in operations research, management science, and analytics to improve operational processes, decision-making, and outcomes through an array of highly-cited publications, conferences and other activities.
My goal was to cover the visualization aspect of unstructured data. As you may or may not know, structured data only accounts for about 20 percent of the stored information. The rest is unstructured data, texts, blogs, documents, photos, videos, etc. While machine learning algorithms have seen significant advancements including analysis of structured data, the tools and processes to visualize the results from these algorithms for the common person have not kept pace. In the article, I argued about the importance of understanding unstructured data as well as enumerate broad guidelines that I follow while building visualization for unstructured data. Overall, the article was very well received by the readers.
Can you talk about the photo management app “Trevi” that you built a while back? How did the idea for the app come about and what did you have to undertake in order to create it?
Kesher: Trevi, launched as Tripsy, was an idea that a friend of mine and I conceived in graduate school. We were passionate about using technology to make travel more social, personal, and fun. Trevi created stories automatically using pictures and videos as well as the location and timeline data. My role was as both a data scientist and the developer of the project. Our biggest achievement was getting featured by Apple in their retail stores as well as getting featured on the home page of iTunes.