From Launching a Successful Startup to Teaching a Class on Entrepreneurship

Computer Science alum Apoorv Agarwal helps new entrepreneurs avoid the common pitfalls of startups

Feb 06 2023 | By Bernadette Young | Photo Credit: Barbara Alper
Apoorv Agarwal

Apoorv Agarwal MS’08, PhD’14 Credit: Barbara Alper

Text IQ, a student startup co-founded in 2014 by Apoorv Agarwal, MS ‘08 and PhD ‘16 in computer science, is one of the few technology startups that became profitable in a year. The startup was acquired by Relativity, a global legal and compliance technology company, in 2021. 

TextIQ uses AI technology to identify sensitive data and prevent high-stakes legal and compliance disasters with a broad range of customers, including Fortune 200 companies, tech giants, and government agencies. The startup is based in New York City with additional offices in San Francisco and Vancouver.

Agarwal co-founded Text IQ while still a graduate student, and his team joined a number of entrepreneurial programs at Columbia Engineering. In 2015 they received an Ignition Grant and then were recruited to go through the NSF National I-Corps program, where they conducted 100 customer discovery interviews. 

Joining Columbia Startup Lab

After receiving his PhD, Agarwal joined the Columbia Startup Lab as a member of the group’s third cohort. Spending the year 2016-2017 at the Lab was a pivotal one for him as it enabled him to hire top engineers to build a product that turned TextIQ profitable in early 2017. 

“It’s been a crazy ride and the Columbia Startup Lab gave us that jumpstart,” said Agarwal, an expert in artificial intelligence and natural language processing. “Recruiting became very easy. When hiring an engineer that Google and Facebook wanted to poach, we could offer a great environment and a great workspace.”

Text IQ’s proprietary software

Text IQ’s proprietary software efficiently and meticulously analyzes big data to identify sensitive, compromising, and privileged documents. It builds AI applications for identifying sensitive information to help organizations reduce their latent risk. It also combines social network analysis and natural language processing to extract the structure from unstructured data, mapping language and people into a Social Linguistic Hypergraph. Text IQ’s mission is to illuminate the world’s dark data so its customers can make decisions based on full understanding. Under Relativity, Text IQ will continue to offer its core products for privacy, legal, and compliance teams, as well as its unconscious bias detector, which is part of Text IQ's AI for Good initiative.

With the successful launch of Text IQ, Agarwal has learned a thing or two about the heartaches and triumphs of running a startup. He knows firsthand that finding the right business idea and creating a startup is a challenging feat. The perfect storm of opportunity, access to the right people, and determination can help launch a startup. But, even with the backing of venture capitalists and investors under the best market conditions, there is no guarantee that your startup will even be profitable, let alone become a unicorn company. 

“When you are in startup mode, you are looking left and right for that silver bullet that will make you successful,” said Agarwal. “And juggling the many responsibilities of running a business for the first time can be really overwhelming.”   

Common startup questions

Over the past four years, entrepreneurs have asked Agarwal about his startup experience. He found himself answering the same questions over and over--what are venture capitalists looking for, how am I supposed to pitch them, how do I generate and validate ideas, how can I build out my team and hire amazing people. Inspired to pay it forward and working with computer science faculty, he created a class that answers many common questions while providing recipes for avoiding failure.

When you are in startup mode, you are looking left and right for that silver bullet that will make you successful. And juggling the many responsibilities of running a business for the first time can be really overwhelming.

Apoorv Agarwal
MS ‘08, PhD ‘16, computer science

Offered for the first time in fall 2022, Agarwal’s seminar, "Topics in CS: Building a Successful Startup,” distills his experience. It is offered to undergraduate and graduate students and hosts weekly guest speakers from his network of venture capitalists, investors, and unicorn company founders. He noted, "One of the key things first-time entrepreneurs miss is a network to tap into. I hope this class enables students to make crucial connections and expand their network." 

TextIQ founders Apoorv Agarwal and Omar Haroun

TextIQ Founders Apoorv Agarwal and Omar Haroun. Credit: TextIQ

Q&A with Agarwal

Columbia Engineering recently caught up with Agarwal to discuss the new class offering and his top tips for entrepreneurs.

Q: There are a lot of classes about entrepreneurship. How is yours different?

I found myself repeating the same advice for entrepreneurs. Several of these entrepreneurs had MBAs but seemed to be missing the real-world feel of getting a startup off the ground. That, plus the need to structure my thoughts and pen them down, motivated me to build this course from the ground up. 

More than 15 guest speakers have talked to our class, and we held a couple of panel discussions, too. The speakers have been very excited to present and interact with our students. These students are super sharp and often throw questions that have pushed speakers to think critically and expand their thinking. Speakers unanimously enjoyed the questions that students asked. 

These investors are trying to figure out how to get access to talent and entrepreneurs earlier and earlier in the process. The earliest they can access them is when they are still at school. Investors have found it helpful to market their brands and connect with budding entrepreneurs, which aligns nicely with my goal of connecting students with entrepreneurs, operators, and investors.

Q: You had the legendary seed investor Mike Maples at a recent class, who said that people shouldn't think about an idea for a startup; instead, the idea should come to them. Do you share this point of view?

Yes, absolutely. For the most part, startup ideas are illogical and irrational. If they were logical and obvious, someone would have already done it. A successful startup is like a mutation that accelerates evolution. It is only in hindsight that the idea seems logical and rational. 

We are wired to think logically, so how do we develop an idea? How do we rig mutation and evolve faster? Mike Maples' long and successful investment career has made him believe that successful startups come from living in the future and that the future is already here; it's just unevenly distributed. He likes to invest in entrepreneurs who are either living in the future themselves or have found someone who is living in the future. 

Q: The idea for Text IQ came from your PhD thesis. How did you transform it into a business?

My PhD thesis focused on extracting social networks from unstructured data such as 19th-century British novels, emails, social media data, and other forms of unstructured data. Natural language processing and machine learning are widely applicable, and every industry has an unstructured data problem. 

Text IQ came from customer discovery or the process of discovering customer pain points. I had never dealt with lawyers before or knew about their pain points. We decided to solve a problem in the legal arena because part of the inspiration for the solution came from my PhD thesis. 

The technology could go through unstructured data such as novels and extract characters in the story. And based on how characters interacted with each other (the language they used and the way they interacted) plus who they interacted with, the algorithms could infer the roles of characters in the story (protagonist, influential, powerful, etc.), their relationships with one another, and how these changed over time. Applying this to emails helped us find people's roles at an organization where titles can be misleading, enabling us to identify fraud, code words, misconduct, etc. Being able to infer such characteristics is important in the context of litigations and investigations. Therefore, our first-use case was geared toward the legal arena. Having said that, we were never an e-discovery or legal tech company. Our vision was to be an unstructured data company that could mine actionable insights from the gnarly and hard-to-deal-with unstructured data.

Q: Based on your experience with creating Text IQ and running a startup, what is your advice for people who want to become entrepreneurs?

I only have two. The first one I got from Steve Blank of Stanford, who said, "Don't make assumptions. Create hypotheses, leave the building, talk to experts, and validate them."

Building a startup is the hardest thing you will ever do. It's inhuman – both physically and mentally. For first-time entrepreneurs, you'll have to drop your life and work 80-100 hours a week (at least for the first two years). If you are not up for that level of commitment, think twice before starting a company.  

Stay up-to-date with the Columbia Engineering newsletter

* indicates required