Release Details

Meet Holly Levanto and Alison Smith

Vice President Holly Levanto was a surface warfare officer with the U.S. Navy. Principal Alison Smith started her career as a business and financial analyst.

Despite their varied career paths, today they share a common pursuit of applying Booz Allen’s industry-leading AI capabilities to U.S. federal government and defense missions. Not only that, but they were also both recently named "Leaders in Technology” by Consulting Magazine.

As an AI adoption lead, Holly oversees the delivery of AI-powered solutions to U.S. defense organizations: simplifying data collection and analysis, transforming logistics and predictive maintenance, supporting warfighters at the tactical edge, and more.

“We’re helping organizations think about their AI maturity, the data they use, and the analytic insights they need, as well as the talent within their organizations and their leadership’s messaging around using tools like machine learning models to enable decision making,” Holly says.

As Booz Allen’s generative AI lead, Alison helps federal agencies see “the art of the possible” with technologies like generative AI.

“Federal use cases are not one-size-fits-all, nor are emerging generative AI applications,” Alison says. “We help agencies figure out how to integrate these technologies into their operations while understanding their risks and limitations.”

Read on for firsthand insights from these AI experts. 

From Driving Ships to Driving AI Adoption

After Reserve Officers’ Training Corps (ROTC) during college, Holly’s first posting in the Navy was on an amphibious ship, the USS Gunston Hall. “It was the best choice for me personally,” she says. “I was excited about being on amphibs, and I like to be an enabler, involved in mission support.”   

A rapid and wide-ranging on-the-job education followed: fixing diesel engines, fighting fires, training officers, advising the commander as the navigator of USS San Antonio’s maiden deployment, and more.

“One of the great things about being a junior officer is you get to learn so much about so many things quickly,” she says.

Even so, Holly wasn’t initially optimistic about her post-military career prospects. “I thought the only thing I knew how to do was drive ships,” she recalls. “But I was wrong. I was a project manager. I designed training. I’d worked on strategy. I’d analyzed solutions.”

Luckily for Holly, Booz Allen had a junior military officer training program that put these skills and more to work through digital transformation programs for defense and military health clients.

This multifaceted background comes in handy in Holly’s current role working with defense clients to implement AI solutions. “It’s not just setting up a data platform,” she says. “It’s making the data readily available, working on the tools, and getting an organization to trust these tools.”

From Economic Modeling to AI Models

Alison’s educational background includes political science, economics, and an M.B.A. But early in her career as a business and financial analyst, these interests evolved. “I started playing around with economic modeling and learned different programming languages to better address data analysis needs, especially as the data formats and volume grew,” Alison says.

This led to research and evaluation engagements in then-emerging fields like natural language processing (NLP). “Iteratively and incrementally, I was able to take the technical and analytical aptitude that I had gained from building macroeconomic models and apply it to a separate problem set,” she recalls.  

She learned NLP and other aspects of AI on the job, and the “almost-hacking approach to problems” she developed has proved valuable. “Both practical and theoretical perspectives are important, especially when you’re on a team solving really big problems,” she says. 

Identifying and Implementing the Right AI Tools

“Wherever there is a challenge, there is an opportunity.”

- Alison Smith, Generative AI Lead

One problem that Alison is helping federal agencies solve is turning an AI vision into an operational reality, starting with picking the right tools for the mission.

“Federal agencies, like most enterprises, are grappling with how to integrate large language models (LLM) into their operations,” she says.

Open-source models are more customizable for targeted applications than commercial LLMs. They’re also easier to deploy in edge applications and secure environments, and they represent the contributions of developers from myriad backgrounds. “This diversity fosters innovation,” she says.

Open source also allows for a more comprehensive study of a model’s limitations, like bias, relevancy, and potential degradation. “Open-source LLMs offer more control to address the government’s unique concerns regarding security, compliance, and privacy,” says Alison. “They can provide the necessary balance between advanced functionality and compliance with stringent governmental standards.”

Yet open-source models can also potentially require more technical capacity and higher development and support costs. Alison’s team helps government agencies evaluate the tradeoffs.

Addressing AI’s Complex Questions

Another challenge: LLMs develop through training, which requires chips, power, and lots of data.

How can these obstacles be overcome? On the data side, some solutions include creating synthetic data from models, improving data selection and curation methods, and developing model training methods that are less data-hungry.

“Wherever there is a challenge, there is an opportunity,” says Alison.

Transformative technologies also spur questions. For example, how does the concept of open source translate to generative AI? This is an area Alison has explored extensively through her work with chatbots and ChatGPT.

“The open-source ethos has long been at the heart of software innovation, championing transparency and collective advancement,” Alison explains. “Yet AI is inherently more complex, less interpretable and explainable, much more computationally expensive, and has many unknown risks.”

From this perspective, she recommends the following question: “What aspects of openness—such as access to source code, training data, and architectures—are most important to your organization, and why?”

A Matter of Technology and Trust

Holly’s AI work similarly deals with questions and answers—specifically, empowering the federal government to make decisions quickly, “moving at the pace of data,” Holly says.

AI can play a powerful role throughout the process, such as collecting vast amounts of data and automating its verification.

But collecting the data is only one part of the picture. “You also have to get the organization to know that their data is authoritative, quality data that it is being governed in the right way,” she says.

That’s where the other part of her team’s job comes in, “bringing clients along on the journey so that they're comfortable with the technology and they trust what's there,” she says.

Another important part of the journey is truly understanding what the data means. Holly explains that if a healthcare organization is tracking safety events, for instance, an increase in the numbers can actually be a positive trend. “It means more people are reporting these events,” she says.

Harnessing Data for Decision Advantage

“I saw how important it was to be able to use data to understand and measure progress and make decisions. And this data is even more important today.”

- Holly Levanto, AI Adoption Lead

As with Holly’s naval experiences and digital transformation work, her work on a client healthcare safety project was a gateway to new horizons. “Assessing the risk in procedures as a part of patient safety really got me into a deeper subset of what data could do,” she says.

Today, Holly plays a leading role in the acceleration of AI adoption by Department of Defense mission commanders worldwide. Her team’s projects include harnessing data analytics to support warfighters and the military’s Combined Joint All-Domain Command and Control (CJADC2) approach to decision making.

“CJADC2 requires seamless data connections between sensors and shooters across services and domains, and this massive amount of data will rely on the power of AI to succeed,” she says.