The Optimizer: Using operations data to make the world run better

The schoolchildren of Hanover, New Hampshire, were caught in a bind. Every morning, they woke early for long bus rides along winding rural roads and through gridlock in the town center, arriving addled and exhausted before class. To help, some parents began driving their children, but with every passenger car, traffic increased and buses were further delayed, creating a vicious cycle. 

It's the kind of problem most academics might pass up in favor of larger and more visible systems. "When I say nobody was looking at this, I mean literally nobody," says Vaze. "No papers, nothing." But it is precisely the kind of problem Vaze likes to sink his teeth into. 

Its solution could transform US transportation and education systems. The national school bus network is the largest transportation system in the world, with more than twice as many rides as all other US public transportation combined. "Also, the length of a student's school bus ride is related closely to health and well-being, including physical and mental health and academic success," Vaze says. "It's a potentially massive impact." 

Vaze first tackled transportation engineering problems as an undergrad in Mumbai and continued as a master's student in transportation at MIT. There, he discovered a whole new field in operations research, which deploys a variety of mathematical tools to find the optimal solutions to complex problems. "I took classes in optimization, probability, game theory, and statistics—and realized there was this whole world with the ability to solve some of the nastiest problems humans face," Vaze says. "If you do it well, it's almost like a free lunch—it requires mathematics and algorithms, but you're not inventing new materials or pouring money into new resources." All you need is a change in approach to reap the rewards. 

In the case of the school bus problem, Vaze worked with current PhD student Prabhat Hegde Th'21 to develop a classic mathematical model that first figured out the locations each vehicle should visit and then the best order to visit, a method they called a "cluster-and-route heuristic." Solving those optimization problems in sequence enabled the pair to cut bus time an average of 40 percent, reduce car trips by around 15 percent, and turn a vicious cycle into a virtuous one. They look forward to seeing how Hanover school district administrators will implement their suggestions. 

While a victory for the schoolchildren of Hanover, the true value of the solution stretches beyond town lines. Some 30 to 40 percent of US school districts are rural, facing challenges comparable to Hanover, and could benefit from comparable route optimization. That fact allows Vaze to have his free lunch and eat it, too: "I'm creating a blueprint for many places while already applying it to one. We all hope people will find value in our work 100 years from now, but it's nice to have some immediate validation at the same time." 

PhD candidate Pushpendra Singh (left) works with Vaze in his Operations Research Group. (Photo by Rob Strong '04)

PEOPLE FIRST, MATH SECOND 

Through the years, Vaze has applied his techniques to a wider array of problems, from transportation to energy and health—a focus that was pleasantly and unexpectedly validated on a recent trip to Walt Disney World, where an exhibit named those three areas as the greatest challenges of the future. "I've been fascinated with these things for quite a while now, so it was an amazing validation for me," he chuckles. In each of these areas, he says, there are three major obstacles. 

The first is the "needle in a haystack" quality of finding the perfect optimization solution. For one problem, he once calculated, there were 10⁵⁰ possible solutions, "roughly the number of atoms on Earth." Of those, there were maybe a few billion that were strong enough to consider, but only one that was truly the best. 

Second, most problems are missing some of the data necessary to find a solution, requiring a method to deal with uncertainty. 

And third, because many problems have multiple stakeholders, what defines the "best" solution can be subjective. 

That's why when Vaze starts investigating a new problem, he doesn't begin with math but with people. "You ask them what the problem is, what they think of as the objective, what are their constraints or limitations, and what levers they have to pull," Vaze explains. Based on those answers, he and other researchers can determine the best tools—from classical modeling to artificial intelligence—to solve a problem. 

In another transportation project, Vaze was trying to optimize flight routing in the face of disruptions due to weather, equipment malfunctions, or staff shortages. 

"Airlines have designed this incredibly large network over months of planning, and they need to redo everything in five minutes," Vaze says. In those circumstances, a time-consuming classical calculation to get the best answer isn’t feasible. Instead, Vaze and students Navid Rashedi Th'25, Nolan Sankey '21 Th'22, and Keji Wei Th'20 used AI to get a timely answer that was "good enough." 

By training a machine learning algorithm on the most likely areas to face scheduling disruptions, the engineers could prune calculations and focus on the most probable solutions. In a test with a major US airline, they found that AI came within 1.5 percent of the most optimal solution. At the same time, the solution saved the airline $400,000 a day compared to existing techniques. "If you can make a small improvement, the overall impact can be huge," Vaze says. 

EXPANDING APPLICATIONS IN ENERGY AND HEALTH 

Two years ago, Vaze saw an opportunity to apply his operations research to the field of energy management and international peacekeeping. Victoria Holt, the director of Dartmouth's Dickey Center for International Understanding, invited Vaze to partner with the center on a project called "Powering Peace," which aims to overhaul energy systems of United Nations peacekeepers in war-torn countries. Currently, such missions rely on diesel generators, which not only have harmful environmental impact but are also susceptible to sabotage or extortion by violent groups, with supply lines stretching hundreds of miles. The project aims to replace generators with alternative energy grids that could both help the UN and create infrastructure for local communities. 

"I knew nothing about it, but it sounded like an important problem," Vaze says. "I came back with a notebook full of ideas." 

Dartmouth participants of the 2026 "Powering Peace" conference in Nairobi, Kenya (l to r): PhD student Lilly Yang, Eric Stambler '89, Professor Steve Peterson, Victoria Holt, Vikrant Vaze, Alejandra Victoria Carrasco Alayo '25, and PhD student Ryan Proulx '25.

Holt appreciated the enthusiasm that Vaze and graduate students Lilly Yang, Siqi Ke, and Ryan Proulx Th'25 showed in approaching the challenge from an operations research perspective. "I've spent most of my career addressing issues from a policy perspective, and it was exciting to bring the rigor of engineering analysis and decision-making models that are heavily based in data," Holt says. "The modeling seems fresh and new to anyone working in the field." 

From the data they examined from Somalia, South Sudan, and Central African Republic, team members knew funding would be a constraint. They developed a model that would yield significant energy improvements without breaking the bank. "We were really trying to push the envelope up to where you could make improvements without exorbitant costs," Vaze says. "Fully getting rid of diesel isn’t necessarily a good idea but getting rid of 95 percent of diesel is in the realm of the pragmatic." 

In addition to the Dickey Center, Vaze has pursued collaborations in health and medicine with doctors from the Geisel School of Medicine at Dartmouth. In one recent effort, he worked with Dartmouth clinicians to explore a machine learning algorithm to diagnose a genetic disposition to high cholesterol, which can cause complications if left untreated but can be effectively managed with medicine. 

Although researchers had already developed an AI model that could examine electronic health records to determine the likelihood of the condition, they still faced the challenge of coordinating broader clinical reviews of patients that could determine the final diagnosis. Vaze applied optimization techniques to create an effective system for patient outreach, including getting buy-in from primary care physicians upfront and reaching out to them through a medical professional, sometimes multiple times, which increased rates of eventual diagnosis. 

The many opportunities Vaze has had for such cross-disciplinary collaborations stem from the nature of close-knit relationships across Dartmouth, from the arts and sciences to its graduate and professional schools and partnerships with Dartmouth Health. 

"Interdisciplinarity is easier said than done," he says. "A lot of that is very unique." In addition, he credits Thayer's broad-based focus, which allows for a more expansive scope of projects than a narrowly focused mechanical or civil engineering-based curriculum. "Here it's about engineering solutions for the betterment of humanity—to paraphrase Thayer's mission—and if you do that, there is always interest and excitement." 

As an indication of his growing leadership role at Dartmouth, Vaze was recently named executive director of the Master of Engineering Management (MEM) program. He steps into the role previously held by Professor Geoffrey Parker, who, as the current faculty director of the Irving Institute for Energy & Society, collaborates closely with Vaze. Parker admires the thoughtfulness and rigor of Vaze's approach. "These are very complex systems with lots of data, and once you bring in limits on what you can and can't do, that means there are tradeoffs," Parker says. "Vikrant brings a data modeling approach using rigorous optimization techniques that is world class." 

Vaze has also developed a reputation as a great mentor, imbuing students with an infectious excitement about the possibilities of optimization and efficiency in operations research. Prabhat Hegde, the lead doctoral student on the school bus project, transferred from a program in energy and initially wasn't sure about the field. "Vikrant took me under his wing and supported me through the transition," Hegde says. "I really came to love my research projects, and a lot of that was working with him and admiring and appreciating the elements in his line of research." 

As he looks towards a postdoc career in the energy field, Hegde hopes to follow his mentor's example of using complex calculations to improve the world. "He has a deep connection to being useful in the real world," Hedge says. "Pursing those types of opportunities is something meaningful I want to continue doing." 

Meanwhile, Vaze continues to prove it is possible to have major impact with a minimum of resources if you can tackle the problem in the right way. "The truth is right there in front of us," he says. "We just need to find the right solution."