|
Listen to this story
|
Walmart, the multinational retail corporation, operates a chain of stores and warehouses. However, like many large companies, it extensively uses data analytics as part of its business operations. The company deploys data analytics to improve inventory management, understand customer purchasing habits, and optimise pricing and marketing strategies.
To learn more about the roles and responsibilities of a Walmart data analyst, AIM reached out to Srujana Kaddevarmuth, senior director, machine learning & innovation, Walmart Global Technology.
AIM: Narrate a typical workday in your role at Walmart Global Tech.
Srujana: Walmart being a Fortune One company, serves millions of customers around the globe and has around 2.2 million associates working towards creating unique shopping experiences for our customers. The data generated is huge and runs into petabytes at any given time. This humongous data needs to be more manageable, heterogenous, and non-intuitive. This is a feast for most data scientists because they can immerse themselves in data and drive value through their work.
At Walmart, I drive the Artificial Intelligence (AI) Modelling Centre of Excellence that involves building commercial-grade data science products for our Omni retail business and new and emerging consumer tech business and data monetisation space. I get to work with and lead exceptional teams of data scientists, AI experts, and machine learning engineers focusing on building recommendation systems, personalisation systems, and voice conversational platforms at scale, using industry trending techniques of computer vision, natural language processing, deep learning, and probabilistic graphical models.
The focus of my charter is not only to drive innovation but to drive the productisation of AI at scale and generate value for our customers, associates, and company while deploying ethical and responsible AI solutions and mitigating unintended consequences.
AIM: Tell us about your prime role vis-à-vis accomplishing future goals for the company.
Srujana: Companies operating in the data space, like Walmart, are now focusing on democratising data and driving significant value for the business by productising data science. This is a journey to translate the findings from exploratory analysis into scalable models that can power products and involves learning various nuances of deploying models into production systems and scaling them effectively.
As an AI leader for the company, my focus is to build AI capabilities at enterprise scale to generate value for our customers, expedite our revenue growth and standardise our tech capabilities to achieve the long-term vision of helping the company be a thought leader in the retail and AI space and move up the automated decisioning value chain. Democratisation of insights by productising data science capabilities helps us progress in this direction. Productising data sciences allows the organisation to move up the analytics and data science value chain quickly. It can help an organisation achieve scale and automation. It enables the organisation to utilise the scarce data science resource for niche data science efforts, freeing them from mundane, repetitive tasks, thereby keeping the data scientists motivated with the work that excites them the most. This would lead to achieving efficiency in utilising human capital.
With a Fortune One company like Walmart that has humongous global operations, productising data science leads to numerous other benefits of standardisation of implementation across multiple data science teams working across different technology domains and geographies. This helps achieve effective data and model governance and enhances the interpretability and reproducibility of complex solutions, thereby achieving fairness and transparency.
AIM: How do you approach an AI/ML problem and ensure that work goes on smoothly?
Srujana: AI/ML models can be great enablers in solving various business solutions. It is essential to focus on building solutions based on functionality and usage, not just based on academic/research acumen. Sometimes the best solutions are the simplest ones. Sometimes most novel and complex solutions may not be computed efficiently and may also not be economically feasible; hence having a product mindset becomes imperative to succeed in the domain.
A perfect model may not exist in this universe; however, building a model viable enough to account for specific real-life considerations and scenarios without needing major architectural redesign poses a significant challenge in ML deployments. Another challenge that often surfaces in ML deployments is production system failures. Machine learning algorithms tend to get smarter over time; however, if they are not connected to new and constant data feeds, they become irrelevant and degrade in quality. To overcome these challenges, we need to build robust data wrappers around the models because the deviations caused due to broken data feed are very difficult to detect compared to application failures. One of the mistakes that some data scientists in the industry make is thinking of the technology stack after the completion of the prototype. This needs to be addressed by planning the product layout, technology stack, and compute resources from the ideation phase. This helps improve runtime performance and compute efficiencies because, in the production stage, run-time efficiencies and functionality take precedence over model accuracy.
AIM: Did you encounter the ‘glass ceiling’ on your way to reaching where you are?
Srujana: It is a fact that the technology industry is male-dominated with more than 80% of positions held by men (as per women in tech statistics 2020). Women hold only 26% of computing-related jobs, and Asian women save only 6%. So, it is evident that many women face the glass ceiling in their careers. As a woman of colour, I felt the pressure of working extra hard to prove myself and make a mark in the industry. I have been very fortunate to find great mentors and allies, both men and women, who supported and sponsored me on multiple occasions early on in my career.
Walmart is a fantastic company and values diversity and inclusion to the core. The company and ecosystem appreciate diversified perspectives and provide a more significant opportunity for meritorious professionals of varied backgrounds to thrive and grow successfully.
AIM: Being a member of the board of directors at the United Nations Association, could you tell us more about your key responsibilities?
Srujana: As a board member and secretary of the United Nations Association, I focus on using data to help progress United Nations Global Sustainable Goals. The 2030 Agenda for Sustainable Development, adopted by all United Nations Member States in 2015, provides a shared blueprint for peace and prosperity for people and the planet, now and into the future. Through Global Pulse Policy, the United Nations is leading efforts to develop data privacy frameworks for the use of big data, to facilitate synergies for the ethical use of artificial intelligence. As a UN SF chapter board member, I focus on mobilising and educating the chapter members about the value of data and artificial intelligence technologies to create social change. We design, organise and orchestrate various events focused on using AI to achieve socio-economic impact.
