As chief data scientist of Federal Hitachi Vantara, Pragyanmita Nayak leads the company’s work to provide government agencies with best-in-class data analytics offerings. A self-described “passionate data scientist,” she has spent more than two decades in the federal marketplace and has published widely on data, artificial intelligence, and more.
Nayak recently sat down with GovCon Wire for an interview with Executive Spotlight, during which she examined the current state of the AI field and shared her thoughts on the direction in which it is heading. She also outlined the challenges federal agencies face in adopting AI and discussed the impact of this technology on today’s defense landscape.
Read the full Executive Spotlight interview below.
Tell me about the current state of the artificial intelligence market. Where do you see new opportunities in AI and where do you think the market is heading?
I have seen an ever-increasing interest in AI and machine learning. Agencies are trying to use it to achieve their mission and goals and are working to better understand where they are going with it. You can see these efforts in relation to the latest executive order from the White House and with agencies coming up with their own data strategy and even AI and cloud strategies. This is proof that this is not just temporary hype or interest in AI.
That being said, 2024 will definitely be the year we see the value of AI manifest. If we look back to 2023, it’s the year AI came to the public’s attention with ChatGPT and big language models, bringing it into the spotlight for those outside the AI field. I expect that in 2024 we will start to see more interesting applications and look for the next high-value delivery application, algorithm model or AI/ML approach. At one point it was deep learning, then came natural language processing. And after? What will the situation be in 2024?
I also think we will see more momentum in traditional machine learning and small language models as we realize that LLMs cannot be deployed and used everywhere, and that there is a lot of work on great language models that we don’t want to abandon. our IP. As a business, when we try to work with large language models, there are approaches to get around this problem.
AI is used today in every field, whether it is manufacturing, finance, healthcare, retail, agriculture or the public sector, and I I expect this to continue to grow as we derive more value from it. There are other areas in the AI/ML space, such as edge computing and natural language processing, that I think will benefit from much better use of these technologies.
As the use of these AI/ML applications increases, we will see much greater growth in the area of ethical AI, bias mitigation, and explainable AI. These three things will become more important as well as increased interdisciplinary collaboration in different areas.
What role can AI/ML play in improving sensor networks and data processing capabilities, and how is your company leveraging technology in this area?
AI/ML is related to edge computing, or bringing your IT and data closer to where the data is generated and where the processing needs to take place so that it is faster, not are not affected by network bandwidth and connectivity and can meet needs in a distributed manner. It is difficult to put the data back on the kernel. If you want to use this data long-term and into the future to do some type of collective analysis and data aggregation, you’ll need it at the core, so it’s important to filter the necessary data.
Edge Computing mainly concerns these sensor networks. Data is collected in these remote locations which often have very hostile conditions and limited bandwidth. There are different ways these sensors work. Data may be collected per second or at a lower frequency, and sometimes sensors collect data at a higher frequency, perhaps on a weekly basis. AI/ML will help solve this type of complexity, tackling the variety of data a sensor can produce and the varying processing capabilities in terms of understanding the data. This will also help filter the data to determine what is important and what should be retained for the long term.
All of this adds a lot of complexity when working with sensor networks. Where Hitachi Vantara Federal comes in is through our years of experience in information technology and operational technology. Hitachi has its roots in manufacturing, industrial equipment, heavy machinery and consumer electronics – and our expertise and digital capabilities have grown around and in support of this area over the past 60 years. Hitachi’s vertical expertise is ideally suited to the convergence of IT and OT domains, so our experience working with sensor-based data prepares us and prepares us to be able to work with this complexity today and bring value.
What are the biggest remaining barriers to widespread adoption of AI by the federal government, and how do you think we can overcome them?
One of the main obstacles to widespread adoption of AI at the federal level is that the effort has been vast, but each agency is trying to do what it wants. While each agency has different issues to address, there are many collaborations that can take place, and I hope the Executive Order will help achieve this. The second factor is the lack of skilled workforce familiar with the AI development cycle. The government is addressing this challenge through one of the criteria of the decree, which aims to increase the workforce and facilitate the hiring of AI/ML specialists as well as the upskilling and retraining of individuals.
The third is that developing AI/ML solutions is somewhat different from the way traditional software development takes place, in which you have defined requirements or you can think about your requirements in advance before even starting the Implementation. In AI/ML, the nature is often more agile, and as you implement and analyze you will discover new trends, patterns and quirks in your data and certain behaviors will emerge that you don’t know about. were not aware of before.
This remains a barrier because the way the federal procurement process occurs through the RFA-RFP cycle does not currently account for this type of agility. The contracting process should evolve and have certain criteria, constraints, and assumptions that encourage this exploration of the solution space as you go through the process of implementing an AI solution. This would ensure that someone designing an AI solution and going through this cycle is not penalized due to delays that could cause them to miss areas of opportunity.
What types of tools and technologies can organizations use to make their data more accessible and understandable?
Every solution you read will say that we are breaking down data silos, but data silos exist for a reason, and that is the lack of trust between different components of an organization. Some data is sensitive in nature and you do not want it to be widely accessible.
Certainly, data silos can also be caused by technological challenges. Sometimes agencies are not able to work in these different locations where data has been moved or they have had to address some compliance or regulatory need where their data has been siled in these locations, either on-premises , or with one or more cloud providers.
Hybrid cloud technology platforms or anything that gives them the flexibility to work on data holistically is important. What helps in this process is good data integration, data orchestration capabilities, and a data platform that can effectively manage metadata so that you have the metadata or data about the data. This information allows you to search your data corpus more broadly and find relevant data for a specific problem.
One tool in this space that makes sorting metadata more convenient is a data catalog. An effective data catalog makes data more accessible and helps you get more value from your data. When you access your data, you can work with it to create more defined applications and get even more ROI on stored data.
In this process, various concepts and technologies, such as data mesh architecture, are gaining importance. This approach means that each domain is responsible for its data, that it makes its data available to others as data products, and that there is an exchange of data products and metadata across all domains. It’s a concept that continues to grow.
How can we make information easily accessible and searchable in all areas? Knowledge graphs and ontology with a data catalog is something I’m hearing a lot more about as part of the journey to extract more value from data. I expect to see more innovation in this space to make the entire data architecture scalable and resilient.
What is your vision of the global defense landscape? What significant changes or trends are you seeing and how are these factors driving the GovCon market?
The global defense landscape is changing because combat today is, rightly, more electronic. You are always told to think like the adversary, and it has become increasingly important to do so in this form of warfare. You need to be on par with your opponent, if not ahead, in order to have a better idea of what kind of access they have to your network and data.
Social media has been used to spread misinformation, which was not possible before. So we need to pay attention to it. I think disinformation is now one of the key elements of war.
In addition to being ahead in terms of AI/ML technologies, being aware of the different types of data and ensuring that your data is not exposed emerge as key elements of algorithmic and electronic warfare, where the AI/ML is used to obtain so much data. we had to get a head start. Having a more automated form of data mining, ML capabilities and pattern analysis is a huge advantage in this form of warfare. We should use these technologies for both the defense enterprise and the warfighter, and we should do it in a way that is intuitive for the warfighter – you can’t expect everyone to be a data scientist. ChatGPT has gained prominence among the most powerful AI/ML features because it is user-friendly and intuitive. When we offer solutions to the warfighter, they must be sufficiently intuitive. It has to be in their line of operation so that it’s organic for them.
