Lila Ghosh, CEO of OxIntel, on how her experience in Hacking for MoD helped define her future

Editor’s note: This is one in an occasional series sharing the journeys of students who have gone through Hacking for MoD and the other mission-driven entrepreneurship courses The  Common Mission Project oversees. 

By Dr. Ali Hawks

When Lila Ghosh began her master’s programme at King's College in London, she was drawn to H4MoD, the academic programme that is helping to solve national security problems at speed, because of the opportunity to work with actual stakeholders within the Ministry of Defense. She was returning to school after a long period in the workforce, and while excited to be in academia again, she really wanted to maintain a level of practical engagement. For her, the opportunity to engage with real-world problems and create tangible solutions couldn’t be ignored. Little did she know that her decision to pursue H4MoD would define her career and help position her as a leading innovator by the time she completed her degree. 

Today, Ghosh is the founder and CEO of Ox Intel, a growing startup operating in the UK and the U.S. that helps design, develop and deploy technology to improve human-led decision-making. 

She joined me recently for a fireside chat, the first for the H4MoD Alumni Network, to discuss her journey through the class and experience as a first-time startup founder (watch the full conversation here). Here are some of her takeaways and insights: 

In H4MoD, Ghosh and her team worked on a Royal Air Force problem related to measuring and streamlining intelligence. 

“We really lucked out with the team, the sponsor and the challenge that we were given,” she said. 

Her team in short order discovered a problem that extended throughout the RAF – analysts were challenged by the sheer volume of intelligence information taken in, and having difficulty synthesizing it into “a unified information environment,” she said.

At the same time, the team was worried about accounting for subjectivity and bias in their processing and in evaluating information. Their task would be developing a software-based system to enable analysts to manage the information overload and process the intelligence in a way that presented an accurate, useful picture for military leaders.  

How customer discovery leads to strategic direction 

The team conducted extensive interviews, 10 each week, to mine the nuances and intricacies of the challenge. They interviewed intel analysts, intel managers, consumers of intelligence, decision makers, and drew on their own experiences by speaking with professors at King’s and people within the team’s professional networks to gain the broad perspectives needed to understand the problem more fully.

“We started with a baseline assumption that we need artificial intelligence. But our bigger challenge, and what our tool is basically trying to solve, is where can AI be helpful? And how can we make sure that we are not displacing the absolute value that human subject matter experts bring to their specific domains? I would say that understanding the problem is an ongoing discovery process that doesn't stop because that single problem means different things to different people,” Ghosh explained. 

The team spent weeks working backward, talking to various users to identify pain points and examine criteria for evaluating intel from an analyst’s perspective. The most surprising discovery was the sheer disparities in criteria from one analyst to the next. 

“We thought: This is interesting. You have two people in identical roles, both trusted to make judgments in a very non-transparent fashion, with completely opposing processes! How can we bring some transparency to this process? And how can we validate some of the assumptions that are feeding into their judgments so that we can ensure that the ways we validate human sources or any other type of source is associated with reliability and performance? That's when we started thinking that we were going to start from scratch and build from the ground up,” Ghosh said.

The sustained intensive study into the RAF’s intelligence challenges soon began to suggest better practices ahead.

“You're starting to make validations that no one else has made because you're applying these methods in such a pace in such a manner that you get to a point where people have been looking at this problem for a really long time,” without much success, she said. “And in ten weeks, you guys have this enthusiastic team that saying, ‘Hey, I think we have a solution.’”

The team began to develop a scoring system for human sources, and is continuing with a similar platform for open sources. Ultimately, their system will be a new tool for evaluating the reliability of information.

The allure of solving real-world problems 

As Ghosh reflects on her start in H4MoD, she understands that she and her team had a rare opportunity to address a significant problem at a macro level, and one that could greatly impact the free world. This was the crystalizing factor that led her to build her company after finishing the course.

“I have turned down what I would have called my dream job going into King’s three times since this all started. That's another thing that you really have to consider going into building your own business -- you must constantly interrogate yourself about the opportunity cost. For me, it was an opportunity that I couldn’t pass up because it was something I came into with an existing interest in. It was challenging, it was new, and it allowed me to still do the work I’m passionate about,” Ghosh said.