AiThority Interview With Yasser Farra, VP of Engineering at Alegion

Know My Company

How have you interacted with AI and other Intelligent technologies that you work within your daily life?

Outside of the AI solutions, we are creating at Alegion, it seems like everything is augmented by AI to some degree now. Whether it’s Netflix or Amazon recommending what you might like to watch or purchase, Facebook facial recognition enhancing your engagement with your friends, Siri/Alexa interpreting your spoken words, Gmail completing your sentences as you type, or Nest predicting the temperature you like in your house, AI is being integrated more and more in our daily lives, often without us even noticing. It is quite exciting to me that AI is weaving its way into everything we do and making our lives better in many ways.

Read More: AiThority Interview Series with Daniel Clark, CEO at Brain.fm

How did you start in this space? What drove you to become an Engineer?

I was always intrigued by computers and interested in programming from the moment I saw an IBM mainframe at Damascus University. At the time, I was pursuing a Civil Engineering degree at the university because everyone in Syria wants to be an Engineer or a Doctor. After I came to the US, and while pursuing a Master’s Degree in Structural Engineering at UT Austin, I taught myself how to program the Mac in its early days, and started a wonderful career in software development, first as a developer and then as a leader, which led me to where I am today.

How do you differentiate Alegion from other AI-as-a-service/ Data-as-a-service (DaaS) providers?

At Alegion, we’re committed to solving enterprise-scale data challenges. We take the time to understand our customers’ use cases and how best to design the tasks to generate the highest-quality data they need to train their Machine Learning (ML) models. The data scientists who come to us have often tried to label and prepare their own data or have worked with one of our competitors, but quickly got frustrated with the time commitment or quality of data they get. The high configurability and advanced capabilities of our platform enable our customer success managers and solutions engineers to work hand-in-hand with the customers throughout a project and iterate on their workflows to ensure the resulting training data meets their ML objectives. We work with customers in short cycles to deliver usable data to them as quickly as possible so that we can adapt our platform configuration early and then on a continuous basis.

Read More: AiThority Interview Series with Kobi Marenko, CEO and Co-Founder at Arbe

How do you see the raging trend of including ‘AI in everything’ impacting businesses?

There is quite a bit of over-hype attributed to AI at the moment. Many businesses believe that AI will significantly improve a host of processes without fully understanding the commitment and cost it takes to develop the ML models that will generate those outcomes. Strategic decision makers must take the time to understand the current limitations of AI and make sure they put the proper resources behind their AI efforts. Enterprises often underestimate the immense amounts of high-quality training data required for AI to become a differentiator, so we spend quite a bit of our resources educating leaders on what investments are needed to reach that point.

What are the biggest challenges and opportunities for companies working to implement AI?

The biggest challenges for companies embarking on AI projects are hiring and building their data science teams, ensuring that the problems they plan to solve through AI/ML are clear and well-defined, and having a sound strategy around their training data pipeline. With the tech giants such as Google, Apple and Facebook investing heavily in AI, and more AI startups sprouting every day, finding the specialized talent in data science and data engineering has become extremely challenging for IT leaders. Not having well-defined problems to tackle with ML consumes precious cycles in enterprise AI initiatives. Finally, the lack of a strategy to scale the training data as projects move from the proof-of-concept stage to production also stalls many enterprise AI projects. But the opportunities presented by supervised ML are boundless. Whether it’s improving customer experience with Chatbots, eliminating cash registers in retail using computer vision, or strengthening fraud detection in financial transactions via NLP, AI will create immense possibilities and differentiation for those who invest properly in their initiatives.

Read More: AiThority Interview Series with Dr. Massimiliano Versace, CEO, and Co-Founder at Neurala

How is Alegion training data platform beneficial for you? Which set of data scientists are best suited to benefit from the training data Alegion generates?

The Alegion platform was built from day one for quality, scale, flexibility, and automation. By leveraging the most stable and advanced backend, front end, storage, streaming, and infrastructure technologies, we are able to deliver large quantities of high-quality training data to our customers reliably and expediently. Our typical customers are enterprise data science teams who are driving large-scale, supervised learning initiatives and are often sitting on a ton of raw data – images, videos, text, or audio – that need to be converted into highly accurate labeled data in order to train their ML models. These data scientists have typically outgrown their ability to annotate their data, or have hit a wall trying to achieve acceptable data accuracy goals. Consequently, they come to us because of our unique ability to improve their ROI through our custom approach and configurable platform that allows us to service any use case these data scientists are trying to solve. By leveraging highly-configurable task designs, ML-augmented workflows, and automation, we optimize not only the quality of the resulting data but also the efficiency of the humans generating those results. This optimization translates into significant cost efficiencies that we pass on to our customers.

How should young technology professionals train themselves to work better with AI?

The AI revolution is creating significant demand for Data scientists, Data engineers, ML Engineers, and Software engineers with AI/ML experience. For technology professionals seeking to enter the AI industry, there is a huge opportunity if they invest time in acquiring Data science and ML skills. There are plenty of free online ML courses, such as Google’s Machine Learning Crash Course, Amazon’s ML training and certification, and Coursera’s ML course. Many of the online AI/ML and data engineering tools are open source and free to use. And there is a wealth of practical examples and online help to accelerate learning. I recommend technology professionals invest time in learning Python and ML lifecycle tools such as Amazon’s SageMaker, Kubeflow, and MLFlow.

There are those who believe that AI will make traditional software development obsolete. I’m not sure that is true. Regardless, it behooves every technology professional to invest the time in AI/ML Learning because that’s where the most exciting innovations are happening, and consequently where the highest paying jobs will be. The definition of a “full-stack” developer may soon grow to encompass ML and data engineering as core competencies.

Read More: AiThority Interview Series With Scot Marcotte, Chief Technology Officer at Buck

How do you consume information on AI/ML and related topics to build your opinion?

Fortunately, the high interest and hype around AI have generated an immense amount of learning and knowledge-sharing everywhere. I try to stay up to date on the latest developments in ML by subscribing to specific blogs of interest, attending AI conferences, going to community events/meetups, and participating in the weekly lunch and learn sessions we have at Alegion. We have a strong culture of learning so everyone at Alegion is proactive spreading knowledge and sharing what they learn on our internal channels.

What makes understanding AI so hard when it comes to actually deploy them? How do you manage these challenges at Alegion?

There is a huge difference between building ML models on a laptop and making them production-ready. On one hand, the ML models themselves need constant tuning and retraining as they encounter new data they haven’t seen before. Doing that while the models are operating in a production environment requires significant investment in build and deployment strategies that are different from those employed in traditional software development. The biggest challenge is acquiring and conditioning data in production, which requires investment in data engineering tools, technologies, and talent. The biggest lesson companies learn when they go from the lab to production is that the ML model is the smallest problem they have; the data pipelines and ML architecture they have to build around the model is where most of the work lies

Read More: AiThority Interview Series with Dean Abbott, Co-Founder And Chief Data Scientist at SmarterHQ

How potent is Machine Intelligence for businesses and society? Who owns Machine Learning results?

This is a very interesting question because it pertains to some of the hype around AI, both positive and negative, that I mentioned earlier. In my opinion, advanced science and technology are always exciting and scary at the same time. Take for example advances in gene manipulation. On one hand, we now have the ability to repair mutated sequences of DNA in any gene which may cure some genetic diseases such as sickle cell anemia. On the other hand, there is fear that such revolutionary science will be used to create desirable physical traits and talents or to genetically enhance crops and livestock, potentially resulting in unintended impacts on the environment. Machine Intelligence is generating similar excitement about possibilities and anxiety about misuse or unintended consequences. I am a big believer that problems created by advancements in technology can and will always be eventually solved by technology. There is a certain level of regulation and societal norms that must be developed to ensure bad actors do not misuse AI.

Where do you see AI/Machine Learning and other smart technologies heading beyond 2020?

Everything we touch and use will be impacted by AI beyond 2020. We already see it in autonomous vehicles, agriculture, medical diagnosis, shopping improvements, fraud detection, etc. I’m most excited about AI introducing advancements in healthcare to provide faster and more accurate diagnoses, aiding seamless globalization by breaking down language barriers, improving security via early detection of malicious intent, and most importantly providing new tools for businesses to serve customers with better products and services. Further out, what some call General AI, Self-teaching Intelligence good enough to outperform humans in many disciplines is the really intriguing and potentially scary future for which we need to be ready.

Read More: AiThority Interview Series with Stuart Brock, Director at Seal Software

What is the Good, Bad and Ugly about AI that you have heard or predict –

More than anything, we hear a lot about AI causing people to lose their jobs. This is not a new notion. You can go back to the industrial revolution in the late 18th century/early 19th century or the digital revolution in the 1980s and the same things were said back then. There will definitely be a short-term impact resulting from AI replacing humans in certain jobs, especially low skill and repetitive jobs. In my opinion, this will be a good thing in the long term. In the last century, those who lost their manual jobs to industrial automation ended up operating computers that control that automation, which means they gained valuable, cognitively-advanced skills. It is incumbent upon society and those in power to invest in skills training of those whose jobs will be displaced by AI in order to avoid a significant generational impact.

Another concern about AI relates to autonomy. Will AI become so powerful and independent that humans will not be able to control it? That’s a question that those developing advanced AI will have to address. But this is a concern of rule-based automation as well, such as airplane autopilots or programmable logic controllers in factories. There will always be a need for the control and governance that allows humans to override AI decisions, especially when we are unable to explain those decisions.

Read More: AiThority Interview Series with Jeff Epstein, VP of Product at Comm100

What is your opinion on “Weaponization of AI”? How do you deal with the challenge here?

There is evidence that AI is already being used for nefarious purposes. A story broke out recently about the Chinese government using facial recognition to track and control the Uighurs, a predominantly Muslim minority group in China. AI is also enabling governments to conduct cyber warfare more effectively and helping create autonomous war vehicles and drones that can initiate and conduct highly-targeted attacks at scale. There is also evidence of AI being used for typical criminal endeavors, such as AI-enabled, highly-targeted phishing. As with conventional weapons, AI-as-a-weapon has to be discussed and dealt with in the same way we discuss any arms race. In my opinion, this is a serious matter of national security that should be on the front burner for policymakers in democratic governments.

What technologies within AI and computing are you interested in?

I’m interested in the ongoing work in Kubeflow, an open source project centered around Kubernetes. There is a significant promise for Kubeflow to make deploying ML models at scale on Kubernetes simple, portable and scalable. I’m excited about AI-optimized hardware which will enable much faster training of ML models. I’m also quite interested in advancements in IoT hardware and software that’s allowing us to implement ML at the edge, enabling us to perform ML inference locally on devices instead of on backend servers.

Read More: AiThority Interview with Alex Fly, Founder, and CEO at Quickpath

As a tech leader, what industries you think would be fastest to adopting AI/ML with smooth efficiency? What are the new emerging markets for AI technology?

Several industries/verticals are typically early adopters of new technology and AI is no exception. In finance, banks have been adding AI solutions for fraud detection and risk management, and brokerages are adopting AI for automated portfolio management, sometimes referred to as Robo advisors, to augment rule-based portfolio management automation. Large retailers are rushing to develop AI solutions to improve their ROI, including grab and go shopping, theft prevention, personalized curation, etc. Companies using drones are leveraging AI for many use cases, including damage detection and assessment (roofs, cell towers, natural disasters), agriculture (soil condition, rock detection, yield management), crowd counting, and of course, numerous security-related applications, both civil and military.

What’s exciting for us at Alegion is that we get to work on all these use cases because our platform is agnostic to industry and can be used to generate training data for any Machine Learning model being developed in any vertical.

Tag the one person in the industry whose answers to these questions you would love to read:

Bill Gates, one of the smartest and most thoughtful technology leaders of our time.

Thank you, Yasser ! That was fun and hope to see you back on AiThority soon.