💭 My experience as an Applied Scientist Intern at Amazon

I recently completed an internship as an Applied Scientist at Amazon, in Barcelona. Here I want to share my experience, partly as a self-reflective exercise but also in case it’s useful to others who may be considering a science role in big tech, especially to those who like me come mainly from an academic background. I will focus on things that I found different—not better or worse, just different—compared to academia and the positives and negatives (but invariably positives) of the internship. I will also share a few thoughts on the full-time role.

First, a necessary disclaimer. My perceptions and thoughts on the internship and role are obviously personal and biased. An internship can also vary a lot depending on the project, manager, team etc.—all the layers of the onion, although the layers closer to you will obviously shape the experience the most.

Differences with academia

Having had no previous experience in big tech, I had no expectations. Below I’ll mention 3 particular aspects of the work that I found quite different compared to academia.

First, the work involved much more software engineering than science. I was essentially tasked with improving an existing forecast to help Amazon deliver packages more efficiently to customers throughout Europe. While the role suggested the use of machine learning (ML), this turned out to be unnecessary. For one thing, we found much simpler ways to improve the forecast, often having to do with data issues and the complex infrastructure surrounding Amazon’s transportation services. For another, replacing the existing forecast with an ML model would not have been a very good idea, mainly because of interpretability—it is obviously harder to debug a black-box ML model than (for example) a logic-based model, which is critical in a business context. I found this to be fairly in line with the often-heard saying that ML in industry is just “regression”.

Another thing that immediately struck me is how fast-paced the work tends to be. There is almost constant iteration and feedback from everyone related your work, from your manager to teammates to business stakeholders. You get the expectation that one should answer questions and fulfil requests as quickly as possible (subject to priorities). This culture is encouraged by one of Amazon’s leadership principles (more on these later) called “bias for action” which rewards employees who take action quickly. It’s related to two of Bezos’ ideas which he recently talked about on Lex Fridman’s podcast. The first is that most decisions are two-way doors, meaning that you can just go back and take the other door if the first doesn’t work out. The second idea is Day-1 thinking, which encourages one to see every day as Day 1, continuing to innovate and avoiding stagnation. Being used to the often too slow engine of academia, I found the high speed and intensity of the work very refreshing. I don’t know how this aspect compares to other tech giants, but I suspect that the more your work impacts the business, the more it cultivates this kind of culture.

It was also interesting to see how widely information in a big company like Amazon is distributed. This meant that it was important to learn who to ask for information, so much so that Amazon’s operation becomes in and of itself an area to develop expertise in. On second thought, this isn’t surprising given the scale of Amazon, and it’s probably also the case at other big companies. Nevertheless, being used to working in relatively small teams, it was interesting to experience this first hand. The complex management structure of Amazon, including the relationship between scientists, product managers, and software engineers, also makes you think about what might be the most efficient way to structure a big company.

Positive & negatives (but invariably positives)

Overall, I had hands-down a fantastic internship experience, with no major negative. Perhaps the only exception was being based in a different country than my manager which, while we made the best of it, would have no doubt further improved our collaboration.

In terms of the positives, I found the culture of Amazon, and the leadership principles (LPs) in particular, very interesting and useful (that’s how you know I’ve been truly brain-washed 😅). The LPs are a set of high-level principles reflecting the values of Amazon as a company that each Amazonian should strive to apply and is in fact assessed on. Besides the bias for action principle, which I mentioned above, I found the following LPs useful at times: (i) customer obsession, (ii) ownership, (iii) dive deep, and (iv) have backbone; disagree and commit. You can look them up here if you’re interested; I’ll just say that Amazon is fairly unique in their regard for the customer, always trying to work backwards from their needs, and you see this very much in the day-to-day work.

Another positive aspect which is often heard about industry was collaboration. Everyone in the team was super welcoming, supportive and always ready to help and answer questions. Yet another thing that was no short of amazing is the large-scale software infrastructure that had been developed for applied scientists, for example to make the best use of the cloud compute on AWS. Related is the opportunity to work with huge amounts of live data. Another well-known plus—especially compared to academia—is pay, which aligns with industry salaries. And let’s not forget about the chance to make an immediate and non-trivial business impact given the large-scale reach of Amazon.

While my experience as an intern was wholly positive, I am not sure whether I would want to switch to a similar full-time position. In particular, I think there are at least two important considerations to bear in mind for someone coming from academia. First, an applied position like this involves—almost by definition—little if any room for more basic or fundamental research. I saw this in both my and others’ work. For me, while I enjoyed the break from basic PhD research and the opportunity to make a business impact, I am not sure whether I’d want to do this full-time. On the other hand, even for more research-oriented positions in industry, it’s hard to imagine a role that allows for the intellectual freedom that a PhD does—which left me with a newfound appreciation for it. The second aspect of the full-time role worth considering is flexibility in working hours. While this is welcomed, it is often hard to implement in practice in an applied context where many issues are time-sensitive.

Conclusion

To conclude, overall I had a very interesting and rewarding internship experience. While I expected to do more science, in retrospect I am so glad to have contributed to making a significant business impact. After all, I think it’s pretty cool to be able to say that you helped make Amazon deliveries throughout Europe more efficient. The experience was surprising in many other ways, including the company culture, the fast pace of the work, and the high collaborative environment. I learned a lot about teamwork, large-scale software engineering, as well as how to deal with ill-defined and ambiguous business problems. If you are interested in getting some big tech experience, especially in a more applied setting, I couldn’t recommend more interning as an applied scientist at Amazon.