In a world that is increasingly becoming digitalized, businesses are relying more heavily on data analytics to drive decision-making. In this setting, tech giant IBM has secured a firm footing in the domain of data science. With the opportunities that the company could offer in this space, how can aspirants get a leg up on a data science career with IBM?
According to the company’s Asia Pacific Leader of Technical Elite Team for Data Warehouse & AI, Vishal Chahal, demonstrating holistic skills around ML Ops as well as Data Ops can go a long way.
“As a data scientist, experience in handling data Ops has become far more important than just a candidate’s educational background,” he says. “They will need to demonstrate the stack skills where they have dealt with data before. A statistical background will be considered an added bonus,” he adds.
Skills Needed For Data Science Role At IBM
The technical skills that IBM looks for in data science candidates encompasses ML Ops, which includes some of the newer skills, like debiasing and machine learning model runtime management.
“In addition to that, they need to possess adequate skills in the areas of Data ops, data wrangling and domain knowledge, which is essentially a cross section between industry knowledge and applicability of machine learning in those industries,” says Chahal.
Although the company does not overemphasize candidates’ educational background, they need to have a good grasp of the relevant competencies mentioned above. With several platforms abound with machine learning certifications, Chahal feels that that may be a good approach for data science aspirants to upskill themselves.
“These certifications can verify their awareness about various platforms, tools, libraries and packages that are being used across enterprises today, as well as the familiarity or the ability to work with open source or enterprise/vendor-specific tools.”
In fact, IBM also offers code patterns on data science for free, which explores the use of machine learning approaches to different industry scenarios and solution domains.
Data Science Projects Over Certifications?
Although the benefits of certifications cannot be emphasized enough, with changing times, the industry requirements for data scientists have evolved too. While online courses have their place in the sector, today the industry is looking for data science stack skills in developing programs, which requires augmented certification along with hands-on experience of having worked on projects. That is, the overall requirement is dual, and this trend is being observed in the hiring practices at IBM as well.
“If you are starting off your career as a data scientist, certifications will certainly help you establish your skill,” says Chahal. “But what will give you the edge is demonstrating a few projects to prove that you have applied the acquired knowledge and skills,” he adds.
According to him, having published code or open sourced on GitHub on data science, or having participated in Kaggle competitions, would prove a candidate’s credential that they have hands-on experience in different fields of data science. “As an accomplished data scientist, we look for experience of having worked on a variety of projects in the data science technology stack.”
Concurs Sharath Kumar RK, who has been working as a data scientist at IBM for nearly four years. “While recruiters will test aspirants’ ability to solve problems on paper, the prime focus will still be on their understanding of challenges at both a micro and a macro level,” he says.
What Can They Expect Once Hired?
According to Chahal, once hired, data scientists at IBM, while focused on getting insights, have to adopt three important data science related best practices:
- Combine and streamline their data science workflows into Data Ops and ML Ops that are enforceable across commonly used open source libraries: In IBM parlance, this is called the ‘AI Ladder’ approach. It encapsulates the importance of having data workflows streamlined to have meaningful data science output. There would also be platforms in place to have built in mechanisms to have a step-up ladder approach to how they do data science, providing them with the flexibility to adopt AI from any step in the ‘AI Ladder’.
- Ability to deploy their Data and ML Ops across multitude of cloud and deployment topologies: Equally important is the ability to deploy the streamlined Data and ML Ops across any cloud, public, private or hybrid. This will provide a simpler and faster way to build, run, and manage AI models and applications across any cloud.
- Explainability and auditability of models built using the data science workflows: Explainability through data science workflows is a core requirement for streamlined Data and ML Ops. Business imperatives require that enterprise AI platforms provide mechanisms that allow decisions derived through machine learning models to be audited and provide transparency around the data being used to build those models to prevent data bias from creeping in.
Hiring Trends In Data Science
According to Chahal, while the popular hiring trend has seen the recruitment of experts from pure data science background with little or no industry experience in the beginning, this is no longer the case.
“Lately, the trend has moved towards recruiting data scientists with stack skills, including Data Ops and ML Ops, or of data scientists possessing domain knowledge,” he says. “Some companies continue to recruit pure data science experts. However, they do look for additional certification, which proves their ability to work across enterprise-wide platforms or open source tools,” he adds.