Haystack Search, New York, NY
I had an incredible experience this summer, learning the ins and outs of a start-up in the seed stage, where the product is not yet completely viable. I worked at Haystack Search, which is aiming to be the future of physical retail, and consequently, we had a huge goal: provide access to specific products as close as possible to the consumer. To provide an example, consider going on vacation in a new place, let’s say Santa Fe, N.M. You check into the Airbnb and see your phone has only 10% battery, and sadly realize that you forgot a phone charger. So, you search for a phone charger on Haystack Search, and it turns out that there is a corner store just a couple minutes away, with a charger in stock. Currently there is no means by which you might have found that corner store. This is the goal of Haystack: to find the closest iteration of a specific product.
Given there is no way to search the inventories of stores nearby, it is extremely difficult to catalog the inventories of all the stores that don’t have their stock stored electronically. With a proprietary utilization of different pieces of data (My apologies for vagueness—I signed an NDA), Haystack predicts if certain stores nearby have specific products. Much of my work at the company related to the processing of this data, which was often times ugly and difficult to interpret, requiring it to be “cleaned.” One project of mine was to make sure the names of stores were correctly capitalized and to remove their names from unnecessary “Inc.” or “Ltd.” tags.
To accomplish this task, I learned a new programming interface, SQL, which is in high demand in data science positions and ubiquitous in server communications. I wrote a program that combed through millions of store names and made sure their titles were correct. For example, a store should not have a word in all caps, if it is not an acronym. Using concepts like Levenshtein distance, my code updated hundreds of thousands of stores and contributed to the database significantly, allowing us to show more stores, more accurately.
Most of my other projects revolved around the proprietary algorithms Haystack utilizes to make predictions and statements about products. The common theme was the emphasis on working with huge amounts of data efficiently: there is not enough time to go to every single store in this country and provide absolutely perfect information. I had to learn to make compromises and still provide the best, most accurate data possible. To do this, one of the things I had to do was actually call certain stores all over the country and inquire about their inventories and status.
The people at Haystack were absolutely incredible. We had a team of only six people, one of whom was working remotely in California and three of whom were interns like myself. Thus, I worked extremely closely with, actually in the house of, the CEO. My proximity to the company’s decision makers put me in the unique and amazing position of being able to ask questions and get the pertinent answer in seconds. On multiple occasions, I had a full-time data scientist showing me the ropes of the discipline, teaching me the tricks he learned from years in the industry. I was extremely surprised at how crude some methods were, but the fact was that they got results quickly, and I learned that a developer’s time is worth much more than the marginal CPU efficiency improvements to be made after constructing a piece of code.
My degree of closeness also lent itself to learning a lot about the company’s financial situation, and the necessary steps a start-up founder must take to ensure success. For example, I learned about the pitch decks that must be shown to Venture Capital firms and incubators, as well as the financing options available. The finances of a high-growth company have long been a great interest of mine. I saw the crazy number of considerations a company like Haystack must make. For one thing, we had to keep tabs on what huge companies like Google and Amazon were doing in our space, which will provide competition. We almost had a heart attack when Google rolled out a “products near you” feature, before realizing it only looks through some chains.
Haystack is not directly involved in areas of significant concern within society because its goal relates to efficiency; however, if the platform does become a significant success, there are very important societal improvements it will make. As of now, most of the stores with a significant online presence are multinational corporations with huge chains of stores, so when someone wants a charger nearby, the only option that appears online is the closest Target or Best Buy. Haystack can point direct consumers to smaller stores, which many have called the backbone of America. Haystack also has the potential to leverage the huge amounts of data it would inevitably collect about local consumer buying habits, to make important discoveries about consumer behavior.
I would say that this internship taught me a significant amount about leadership and how managing a business relates directly to managing people. As an aspiring entrepreneur myself, seeing the daily relationship management, not to mention value of networking, was extremely valuable to me. As the leader of a company, I learned, it is absolutely vital to define expectations and designate tasks to the correct personnel. Moreover, there is a fine line in work culture between lax and stringent. There’s also a valuable intangible ability to show other’s you are worthy of following, which has a lot to do with the seriousness of one’s demeanor.
Working at Haystack showed me how it is to have a real job. I had a forty-hour work week, eight or nine hours each day, with a two-hour commute in both the morning and at night. I woke up at eight, went to work, and could barely get back twelve hours later. That experience me about commuting to work and maximizing my potential in my free time. The subway ride could be spent reading, and the train on sending out emails. This internship showed me that real work takes real work ethic.
Additionally, my internship at Haystack caused me to consider a variety of new options when I return to Williams. For example, I would like to learn more about data science in a course. My experience combined with the advice of a distinguished fellow Eph—to enter the work hierarchy on the production side, rather than in management where it is impossible to switch—have convinced me to heavily consider majoring in computer science. In terms of my work after college, my experience has affirmed my interest in high-growth technology firms, but I’m still not sure exactly where I want to be.
The Alumni Sponsored Internship Program sets Williams apart from its peers, allowing students like myself to get experience in the hugely saturated college labor market. I would like to thank everyone in the ’68 Center for Career Exploration, especially Dawn Dellea, for managing this program. I’d also like to thank Ben Dean at Haystack Search for providing me with this opportunity and Nikita Bondarenko for being a great teacher. Finally, I would like to thank Williams alumni Peter ’79 and Laurie ’79 Thomsen for their generous support of this internship.