When a Stanford Bible Study Led to an AI Startup

Bible studies at Stanford University sometimes lead to an outbreak of datasets.

Hadassah Betapudi and Elijah Kim met at a Christian fellowship at Stanford in 2022 and got to know each other by leading a Bible study together. Soon the duo—with their backgrounds in data organizing and computer science—was building an artificial intelligence startup.

The two sought to solve a problem: They had heard from students needing a lot of guidance on the essay-writing part of college applications. That mentoring wasn’t available to many because of their financial or cultural backgrounds.

“I see education as a super critical way to directly bless people in a way that closes opportunity gaps,” said Kim.

Neither Betapudi nor Kim felt that they had entered the application process with a lot of guidance themselves; Betapudi was the first in her family to go to college in the US, and Kim was homeschooled his whole life before college.

“Applying to college was a little scary because I lacked a lot of the insight and guidance that I think a lot of students take for granted,” said Betapudi, who was born in Memphis to Indian immigrant parents. “It’s not just about your test scores or your GPA but about presenting who you are holistically to a college admissions committee, which is very different than the way things are done in India.”

This year the duo launched their startup: an AI tool that provides college admissions essay feedback. The founders see it as more like a guidance counselor and editor, not a content producer like ChatGPT. It doesn’t write essays for students.

Called Esslo (a mashup of essay and Elo, the chess rating system that influenced their proprietary algorithm), the AI engine was born after Betapudi and Kim trained it on good and bad essays from students who were admitted to top schools like Stanford. They taught the AI how to generate feedback on those essays. They beta-tested the tool this spring and launched it this summer.

Now Esslo is growing. The founders say the tool has been pulling in 100 new sign-ups a day.

Dustin Nguyen is one, a senior at Bronx Science, a STEM-focused public high school in New York City. He’s planning to apply to 22 colleges, and he’s already applied to 6 to meet early-action deadlines. But he must write essays for all of the applications. His public school counselors don’t have time to help him refine his essays.

Nguyen found Esslo this fall. To use Esslo, students feed their own essays into the model and get detailed feedback on the quality and suggestions for improvement. Sometimes, Esslo can tailor feedback to specific schools or admissions standards.

The tool gives line-by-line comments and scores the essay on traits like “detail” and “curiosity.” Then it gives overall feedback on the essay, followed by “brutally honest thoughts,” which might say something like “The writing in this essay is mediocre.”

After using their scientific knowledge to build the AI engine, the founders incorporated their own writing knowledge into the algorithm. For example, Betapudi edited a Christian journal while at Stanford. The machine tells students to have a “hook,” or a strong opening line, and to “show, don’t tell.” The creators recognized that a winning college essay is different from a typical writing assignment a student might do in high school.

“Technology with humans is always stronger than technology that’s existing in a vacuum,” Betapudi said. “It’s the beauty of a conversation with someone else that’s able to unlock thoughts about yourself or insights that you wouldn’t have had otherwise. I mean, that’s why people love therapy or coaching.”

When Nguyen fed Esslo an essay he wrote about being Vietnamese, the tool highlighted one anecdote in his essay and told him to elaborate on it and make it feel more personal.

“It’s helped me refine my writing,” he said.

Nguyen grew up in a low-income family. Most students can pay a monthly fee to use the tool, but Betapudi and Kim have enough paying customers to be able to give Esslo access to high schoolers who meet certain financial criteria, like being on free or reduced lunch or being at a Title I school.

“It levels out the playing field, especially since I can’t afford private counselors or special programs,” Nguyen said. “Esslo is a really good example of using AI to the benefit of a lot of people, especially underrepresented and underprivileged people and communities.”

That was part of the faith-based motivation for Betapudi and Kim.

“The truly dystopian outcome of introducing AI into education or into the world is that these oppressive regimes or governments will have access or will develop better AI than the good guys—those who fight for the widow, the orphan, the poor, or the lame,” Betapudi said. “I have been gifted through grace the ability to go to Stanford. … As a believer, my charge is then to build tools that not just benefit the top richest 1 percent who is already using this sort of thing but making it equitable and making it easy to access.”

Kim sees helping students as an “extension of discipleship. …… Jesus discipled a few, and then those disciples went out and discipled others.” He has seen the impact of personal mentorship and guidance in his family. His grandfather didn’t know anyone when he came to the United States from Korea, but Kim said someone came alongside him, “poured a lot into him,” and helped set him up to care for his family.

“That is very motivating to me personally,” he said. Education “sets people up in a way that they’re able to not just help themselves in the future but … help others.”

One college counselor sees potential for Esslo to serve those without counseling resources in Mongolia. Khongorzul Bat-Ireedui used to run a private school in the large central Asian country, where she counseled students going to college. She still does this but now from California, where she is in graduate school. She has been using Esslo to analyze her own counseling feedback on student essays and said it helps her see things she might have overlooked or observations from a different perspective.

Bat-Ireedui has a substantial following on Facebook in Mongolia, and she shared the tool on her page. She thinks Esslo will be helpful there because Mongolian public schools don’t have counselors.

“Essay writing itself is hard for an [American] student,” she said. She thinks about Esslo for the Mongolian public school students she knows who have worked hard to learn English and have good test scores but may not have money for a private counselor. “We don’t have schools that teach these things.”

Another college counselor, David Heinemann, has also used it to analyze student essays.

“I was skeptical in the beginning. … I don’t like the use of AI in the college process: the way colleges are using it, the way students are using it, and even the way counselors are using it,” said Heinemann, who works at Vail Mountain School in Colorado. “But … I was blown away. The specificity with which Esslo gives feedback was almost scary. … It just did a good job of doing what I do.”

Heinemann said the machine doesn’t give generic feedback but offers ideas for different directions the student could take the essay. Given the 650-word restriction on a typical essay, he said the feedback sometimes almost gives students too many options. But he thinks that is good for students: they must make decisions about what advice to listen to.

“It’s not like cheating—it’s not giving you the answers or rewriting the essay,” Bat-Ireedui agreed. “It’s giving you the feedback a human counselor would, but it’s picking up stuff that a human counselor might have missed.”

Heinemann thinks the founders could make “a ton of money” off their AI model, but it seemed to him like they were developing it to help people rather than cash in.

Both Betapudi and Kim think about their creative work with AI as reflecting God’s creative work.

Kim works in a lab at Stanford now studying energy efficiency in new computing technologies, and it puts him in awe of the human brain that can do tasks using 20 watts of power when the same task might require an entire data center for AI. Kim can’t build a brain, but he’s always loved building new things and understanding how they work.

“God is the ultimate designer, the ultimate engineer,” Kim said.

The post When a Stanford Bible Study Led to an AI Startup appeared first on Christianity Today.

Translate »