For a researcher, publication in Science is considered a career-defining milestone. For work on human trafficking, Kejriwal says, visibility is vital for another reason.

“Trafficking often thrives because it stays hidden in the noise of the internet,” said Mayank Kejriwal, research associate professor in the Daniel J. Epstein Department of Industrial and Systems Engineering at USC’s Viterbi School of Engineering and principal scientist at USC Information Sciences Institute. “By highlighting this research, we are signaling that trafficking isn’t just an inevitable and tragic social issue but a problem that can be tackled through science and engineering.”

An Automated Detective’s Clue Board

The paper describes nearly a decade of work developing an AI system that turns the fragmented digital traces traffickers leave behind into evidence of sex trafficking that holds up in court. The technology has already helped secure at least six trafficking convictions across the country.

When human trafficking moved online, it became nearly invisible. Criminals could post hundreds of advertisements a day across thousands of websites, hiding behind misspelled phone numbers, emojis, and fake names. For overwhelmed investigators armed with little more than Google searches and spreadsheets, finding victims was like searching for needles in a haystack that kept growing.

Kejriwal saw an opportunity where others saw only chaos.

He was a key architect on a system called Domain-specific Insight Graphs, or DIG, which works like an automated detective’s clue board. “In a traditional investigation, a detective might spend weeks manually linking phone numbers or names across thousands of sex advertisements. DIG does this in seconds,” Kejriwal explained. “For example, if a trafficker posts ads across 10 different websites using different names but the same hidden phone number, DIG’s AI recognizes that connection.”

Hundreds of Millions of Ads, Thousands of Websites

The scale of the problem is staggering. Sex trafficking advertisements number in the hundreds of millions in North America alone, scattered across thousands of websites rather than concentrated in one place.

“When trafficking moved online, it gained the same advantages as legitimate digital businesses: speed and massive reach,” Kejriwal said. “This fragmentation used to be their biggest shield. They counted on the fact that no investigator could manually keep up with so much data scattered across so many corners of the web.”

But that digital footprint also became traffickers’ vulnerability. Kejriwal’s team trained the AI specifically on the tactics used in this underground market. “A typical search engine only finds exact matches. If a trafficker writes a phone number as ‘seven-zero-seven’ to avoid detection, a regular search will miss it,” he explained. “DIG is smarter because we trained its AI specifically on the ‘language’ of trafficking.”

Building a Time Machine for Evidence

What surprised Kejriwal most during his extensive consultations with prosecutors and federal agents wasn’t just that they needed faster searches. They needed a time machine.

“I initially thought they just needed a faster search engine, but I soon learned they also needed a time machine,” he said. “Prosecutors told us they didn’t just care about what was online today: they needed to see (and preserve) what happened six months or a year ago to prove a case.”

That insight transformed DIG into a system that archives and preserves evidence for the long term, even after traffickers delete their ads to hide their trail. In one documented case, investigators used DIG to secure a conviction by showing that a trafficker was posting advertisements depicting a victim while she was lying unconscious in the hospital. The preserved evidence helped refute his defense that the victim’s participation was consensual.

Open-Source Tools and Taking the Fight Global

DIG was built using open-source tools rather than expensive proprietary software, and Kejriwal’s released the code publicly. The decision was deliberate.

“Fighting crime should not be a luxury available only to the wealthiest departments,” he said. “Many agencies on the front lines are underfunded and cannot afford multi-million dollar software. By using open-source tools and releasing our code, we enabled entities ranging from the local police department to a global NGO to access these same capabilities.”

Now, through the Global Trafficking Initiative he co-founded, Kejriwal is expanding the work internationally. The challenge becomes harder across borders, where countries have different laws and levels of resources.

“The challenge becomes much harder when you cross borders because every country has different laws and levels of resources,” he said. “A trafficker might recruit someone in one country, host their ads in a second, and exploit them in a third.”

A Path to a Trafficking-Free World

His article in Science concludes with a bold claim: “A path to a trafficking-free world is feasible.”

To get there, Kejriwal says, requires more than just better technology. “We need legal frameworks that encourage data sharing across borders and social policies that are informed by the experiences of survivors,” he said. “To make a trafficking-free world a reality, we have to stop treating these as isolated local crimes and start treating them as a global network that can be mapped and dismantled.”

After nearly 10 years working on this problem, he’s seen enough to believe it’s possible.

“What kept me going was seeing the technological unfairness of the situation,” Kejriwal said. “Traffickers were using the latest digital tools to exploit people, while the investigators trying to stop them were often stuck with outdated technology. Every time a lead from our system helped an investigator or a prosecutor, it proved that this was a problem we could solve if we just gave the ‘good guys’ better tools.”

Published on February 5th, 2026

Last updated on February 5th, 2026