Four steps to accelerate your machine learning journey

By Swami Sivasubramanian, Vice President, Amazon Machine Learning, AWS,

This is the golden age of machine learning­ (ML). Once considered peripheral, ML technology is becoming a core part of businesses around the world, regardless of the industry. By 2021, the International Data Corporation (IDC) estimates that spending on artificial intelligence (AI) and other cognitive technologies will exceed $50 billion.

Locally, 25 percent of organisations say they are setting aside at least 10 percent of their budget for technology, which includes investments in big data analytics (64 percent), cloud computing (57 percent), Machine Learning and artificial intelligence (33 percent), and robotic process automation (27 percent), based on the Malaysian Institute of Accountants’ “MIA-ACCA Business Outlook Report 2020″. [1] As more companies gain awareness of the importance of ML, they should work towards getting it in motion as quickly and effectively as possible.

At Amazon, we have been on our own ML journey for more than two decades – applying it to areas like personalisation, supply chain management, and forecasting systems for our fulfillment process. Today, there is not a single business function at Amazon that is not made better through ML.

Whether your company is just getting started or in the middle of your first implementation, here are the four steps you should take to have a successful ML journey.  

Get Your Data in Order

When it comes to adopting machine learning, data is often cited as the number one challenge. We found that more than 50% of time spent in building ML models can be spent in data wrangling, data cleanup, and pre-processing stages. Therefore, prioritize investing in the establishment of a strong data strategy to avoid spending excessive time and resources on data cleanup and management.

When starting out, the three most important questions to ask are:

  1. What data is available today?
  2. What data can be made available?
  3. A year from now, what data will we wish we had started collecting today?

In order to determine what data is available today, you will need to overcome data hugging – the tendency for teams to gatekeep data they work with most closely. Breaking down silos between teams for a more expansive view of the data landscape while still maintaining data governance is crucial for long-term success.

Additionally, identify what data actually matters as part of your ML approach. Think about best ways to store data and invest early in the data processing tools for de-identification and/or anonymization, if needed.

Identify the Right Business Problems

When evaluating what and how to apply ML, focus on assessing the problem across three dimensions: data readiness, business impact, and machine learning applicability.

Balancing speed with business value is key. Instead of trying to embark on a three-year ML project, focus on a handful of critical business use cases that could be solved in the upcoming six to 10 months. Start by identifying places where you already have a lot of untapped data and evaluate if ML brings benefits. Avoid picking a problem that is flashy but has unclear business value, as it will end up becoming a one-off experiment.

Champion a Culture of Machine Learning

In order to scale, you need to champion a culture of ML. At its core, ML is experimentation­. Therefore, it is imperative that your organisation embrace failures and take a long-term view of what is possible.

Businesses also need to combine a blend of technical and domain experts to work backward from the customer problem. Assembling the right group of people also helps eliminate the cultural barrier to adoption with a quicker buy-in from the business.

Similarly, leaders should constantly find ways to simplify the process of ML adoption for their developers. Since building ML infrastructures at scale is a time and labor-intensive process, leaders should encourage their teams to use tools that cover the entire ML workflow to build, train, and deploy these models efficiently.

For instance, 123RF, a homegrown stock photography portal, aims to make design smarter, faster, and easier for users. To do so, it relies on Amazon Athena, Amazon Kinesis, and AWS Lambda for data pipeline processing. Its newer products like Videomaker uses Amazon Polly to create voice-overs in more than 10 different languages. With AWS, 123RF has maintained flexibility in scaling its infrastructure and shortened product development cycles and is looking to incorporate other services to support its machine learning & AI research.

Develop Your Team

Developing your team is essential to foster a successful ML culture. Rather than spending resources to recruit new talent in a competitive market, hone in on developing your company’s internal talent through robust training programs.

Years ago, Amazon created an in-house Machine Learning University (MLU) to help its own developers sharpen their ML skills or equip neophytes with tools to get started. We made the same machine learning courses available to all developers through AWS’s Training and Certification offering.

DBS Bank, a Singaporean multinational bank, employed a different approach. It is collaborating with AWS to train its employees to program their own ML-powered AWS DeepRacer autonomous 1/18th scale car, and race among themselves at the DBS x AWS DeepRacer League. Through this initiative, it aims to train at least 3,000 employees to be conversant in AI and ML by year end.

[1] MIA (Malaysian Institute of Accountants) and ACCA (Association of Chartered Certified Accountants), Business Outlook Report 2020, 2020


Please enter your comment!
Please enter your name here