At Sift Science, interns have been an integral part of our company’s culture and growth. So it’s no secret that we’ve been working tirelessly to build a valuable engineering internship program. Our team is committed to building a program that fosters a culture of mentorship, transparency, and collaboration.
Having an internship program isn’t just about training the next generation of technical talent: it’s about giving students the opportunity to work on tangible problems where they can add value. Whether they’re working on machine learning and big data or storytelling and visualizations, our interns are able to make a big, meaningful impact on our company and technology. But don’t just listen to me…check out what some of our summer interns have to say about their experience!
Will Song, University of California, Berkeley, Sift Science Intern, 2017
What did you work on during your summer internship at Sift Science?
I worked on a couple projects. One was parallelizing online scoring to lower latency. For each score request from a customer, we were evaluating models sequentially and unnecessarily increasing latency. By parallelizing model evaluation, I managed to cut down online scoring latency by 30-40%. This was a change in a core code path and something that every customer of Sift would benefit from. In addition, I made a tool with MapReduce to compare the parameters of our models. The tool gives insight into how our models change after each model refresh and also helps engineers debug any modeling issues. This project was somewhat open-ended in terms of how exactly one should compare models, and I really enjoyed reading up on some of the math and having the autonomy.
What knowledge or skills have you gained from your internship?
I learned a lot about the type of problems that arise when applying ML at scale and in industry (very different from the sandbox ML problems I was exposed to in school). Sift is also really transparent on how the company is run. I gained a surprisingly good amount of knowledge about the various business/marketing/sales strategies it takes to run a startup. Sitting in on sales calls really opened my eyes to the concerns of other enterprises.
How would you compare Sift’s internship experience with others?
Sift was definitely the best because the projects I worked on really made an impact. Other companies give you mundane or pet projects and some projects may never even see the light of day.
Vincent Chen, Stanford University, Sift Intern, 2017
What did you work on during your summer internship at Sift Science?
As a member of the Little Data team, I worked primarily on the customer-facing console and distributed automation engine. Throughout the summer, I was able to work on a wide range of projects, from improvements to the workflows interface to changes that powered our internal analytics. One of the larger projects that I shipped allowed analysts to set sort priorities in their manual review queues.
What was the most memorable part of your summer internship at Sift?
One of the most memorable moments from the summer was during the week-long summer hackathon. I was working with several engineers on the Abuse Products team, and we came together from different contexts to tackle some early assumptions that were made about automated labels for machine learning. This was a wonderful example of collaboration — having worked on the automation before, I gave context into how customers used our fraud engine, while the members of the modeling team offered insights into how certain automations were fed back into the learning process. Ultimately, this experience was very rewarding because of encouraged collaboration and learning from all sides.
How would you describe the culture and people at Sift?
Sift prioritizes a healthy, collaborative, and challenging culture. If it gets that right, other things fall into place. Sift was one of the few places where I felt pushed to grow by people who wanted me to grow. I was surrounded by individuals who were compassionate and supportive, and it certainly feels like other things are falling into place as well.
Tim Wang, University of Maryland, Sift Intern, 2017
What did you work on during your summer internship at Sift Science?
My main project was working on Machine Learning models that learn from text. The project was extremely fun and rewarding, and it allowed me to learn a lot about the Sift Science code base. More importantly, it was the first Machine Learning model I built purely from scratch and it was the first that had to be able to handle the scale of data Sift has. Finally, my project kind of pioneered Sift’s foray into the next generation Content Abuse detection product and I’m glad that I was able to lay down the foundation for that.
Open forum! Anything else you would like to say regarding your experience at Sift?
Overall, it was amazing. If you intern at Sift, you will get a hands-on project that directly appeals to your interests and will make an actual difference in the real world. Furthermore, given that Sift is starting to grow at a tremendous rate, now is a great time to come aboard.
Irving Chen, University of Washington, Sift Intern, 2017
What did you work on during your summer internship at Sift Science?
I worked on Sift’s new Account Takeover (ATO) offering. My two main projects:
1. Revenue estimation: predicting the revenue that comes from ATO customers. Revenue estimation is used on the business side for strategic planning, tracking burn, etc. I implemented this in two parts: a recurring MapReduce job to compute the latest usage statistics relevant to our ATO pricing plans, and an estimation component that combines historical usage trends with the latest usage statistics to estimate future usage (which is then fed into our invoicing system to calculate actual $$$ numbers).
2. Accuracy metrics: enabling us to measure the accuracy of our ATO product. ATO is still pretty new, and only recently has Sift acquired enough “ground truth” (“decisions” from customers) to start answering the question “how good is our ATO product?” I wrote a MapReduce job to pair up ATO scores (0-100) for login events with the actual decisions (good or bad) on these login events from customer analysts, and then connected this data to some existing Jupyter notebooks we use for model evaluation. I also picked up smaller projects here and there as they came along, since I like working on many things at once.
What knowledge or skills have you gained from your internship?
On the technical side, I got a lot of experience with Hadoop/MapReduce. MapReduce in principle is pretty simple but when it comes to actually using it I found that there were all kinds of pitfalls (and I fell into a bunch of them…). I also worked with a bunch of other technologies to varying degrees, including GitHub/git, IntelliJ, tmux, HDFS, HBase, Mongo, Spark, Jupyter, and Crunch.
Do you have any previous internship experience? If yes, how would you compare them to Sift’s?
I interned at Microsoft back in high school and more recently at Google, in Mountain View and in New York. I think internships at larger companies are more of a hit-or-miss kind of situation, since there are so many teams and so many different projects, and my prior internships tended to be a “miss.” Things that made Sift a “hit” for me:
– Way easier to ask questions/get help on cross-team things. Everyone is in one big Slack team and/or a few steps away. (Sometimes it was a journey at Google — literally, biking across campus — to talk to the right people.)
– Better thought-out projects. At Google, my projects weren’t very related to what my teams were doing, and I spent a lot of time digging through documentation on my own. My projects at Sift have been more directly relevant to my team’s work.
– More accessible codebase. This is more of a personal thing but I like to pick up small work items here and there and that’s easier to do at Sift, where systems aren’t too complicated (yet), versus at Google where even side projects take weeks or months.
Peter Lu, Stanford University, Sift Intern, 2017
What did you work on during your summer internship at Sift Science?
I worked on restructuring the offline UsersPer counting system used to generate valuable features for the ML pipeline (and with implications for ATO). The current UsersPer system is flawed for a few reasons, the biggest being that counts are generated from only about 5% of our offline database. My changes will bring this to 100%, hopefully generating more accurate counts for model training.
What knowledge or skills have you gained from your internship?
I learned about how to contribute code within a large complex system. Sift is unique that it is a company developed enough to have a fully polished proven product, but small enough to where much the code still remains exposed to the developer. Because of that, writing code for Sift involves working across the stack and across teams to make sure everything integrates smoothly.
I also learned about the organizational structures necessary to help all Sifties synergize and improve the core product.
What was the most memorable part of your summer internship at Sift?
Driving up to Salem with fellow Sifties Katherine and Patrick to watch the eclipse was unforgettable. On the way we picked up Jason (CEO Jason!) which was awesome. Nothing bonds people together quite like a 14 hour drive through ridiculous eclipse traffic. Except maybe watching the eclipse itself.
Reggy Long, Stanford University, Sift Intern, 2017
What did you work on during your summer internship at Sift?
- Adding URL feature extractor
- Using sequence learning to improve our fraud prediction scores
- Changing the feature extraction pipeline infrastructure to save computation time (up to a few hours)
Would you recommend a friend to intern at Sift next year? Why or why not?
Yes, because you can have a lot of impact (since the company is still small) in a successful company.
