Meta (Behavioural)
Tell me about a time you had a disagreement?
So what feels like now an in another life time, I used to be a specialty coffee technician, so I worked on commerical coffee machines in the specialty coffee industry. One of the many ways we fine tune the coffee is controlling the kind of water that is used for the coffee. So to get started with that you would need a filtration system, and the water filtration industry particularly for specialty coffee is very niche and limited. However, my coworker who was there before me had reverse engineered the filtration system for half the cost, so we could then pass that cost savings on to the customers. It was very brilliant of him, and so I was also apart of setting up these filtration systems when I came on board.
Throughout building these systems, I realized there was some upgrades we could do that would decrease cost by an additional 15% and increase production by 30%. So seemed like a win-win, however when I presented this idea to him and our manager who also was one of the co founders at the time, he was very sceptical about the proposed changes and when we finished that discussion it felt almost like he was saying it is already working well there is no need for further progress. Our manager had no technical experience so he defaulted to him, as he had been there for longer, and also like I mentioned he was a very smart individual. I just had never considered that this project was what got him his recognition at this company and by potentially proposing these changes it might minimize his impact.
So I could have just left it at that, however I felt it was important to follow through with these changes as it can lead to a massive increase in business impact. I pulled him aside afterwards, and spoke to him to get all his opinions on why he disliked the proposals, and some of them had some significant weigh to them. So over the upcoming weeks I continued this open dialogue and collaborated very closely with him to address these changes, and having open conversations for further improvements. Finally at the end of those discussions he was completely on board and through our open dialogue we had an even bigger business impact than I first proposed. Our new proposals had a 30% decrease in cost and 50% increase in production. We did some testing and then implemented these changes, and it became a further staple for the business to distinguish themselves from our competitors.
Tell me about the project you are most proud of?
I’m particularly proud of my face mask detection project, in which I used neural nets to train an optimized model that was able to detect in real time whether an individual was wearing a face mask or not.
I considered this an important problem to solve because this was near the time when lockdowns were being lifted and face mask were required in large public settings. However, compliance is difficult and in large venues it was hard to enforce this because it required so many personelle to basically search and monitor. So I had to think of a better solution to make compliance more feasable particularly at scale, and that was when I had the idea of detecting compliance issues using something along the lines of a CCTV to highlight particular individuals and zones to dispatch personelle to so they don’t have to search aimlessly.
Before this I had never encountered anything in machine learning before, I had no acadameic experience with it either. So it first required just digesting in a lot of information to demystify neural nets, setting up the tech stacks so I can train locally and make sure I understood the problem well enough to be able to gather data and optimze for that problem. So I started by contacting one of my professors who specializes in image recognitions problems and just bouncing off my idea and asking for some resources. He loved the idea and he recommended a few articles and books which I used as a mini crash course pre-requiste before staring to tackle the problem.
I then began building the MVP, was able to train a few models with low accuracies, and then spent time optimizing the hyper parameters and gathering more data. In the end I was able to get a validation accuracy of 93.45% which I found sufficient for the problem I had defined in the beginning.
Later on when applying for my first internship, the CEO of the startup I applied to did a very deep dive into my project, the commits, the particulars and was very impressed with it, particularly taking an idea to practical applications. That’s how I was able to get my first internship as an ML engineer, and so is one the projects that I’m most proud of.
Tell me about a time you failed or made a mistake?
So at my current intership at SAP after the onboarding and a couple of starter tickets I was given my first more complex ticket. This was given after discussing with our APO, PO, and Tech lead. I very enthusiastically made a plan, got it signed off by our PO & APO, and then began implementing my proposal. However, one of the ways to gather resources in our org was done in a particular way, which in my opinion was a bit slow, so I suggested an alternative to speed up this resource gather. However, at the time of this suggestion our Tech Lead was on vacation, and when I mentioned this to the APO & PO, they suggested I should wait until he comes back before continuing forward.
I understood, however I thought maybe I can go forward with implementing my ticket with the idea that the alternate solution would be taken. So I did that, however when our Tech Lead came back he was not particularly on board with the alternative suggestion I made, he had some of his own compelling reasons, and technical debt which I hadn’t considered, so this meant I had to redo the implementation of my current ticket as the current alternative would no longer need to be in place. I apologized to our APO & PO, and told them I should have waited and followed their suggestion earlier. Ultimately however it was a very positive discussion with the tech lead, as he understood my alternative solution and we even made some partial implements for faster grabbing, however I also was able to really understand the full scope of the org and the suggestion I was asking after I spoke with him. So this made sure that moving forward I would ensure all parties are algined in our org or cross organizationally before beginning to implement a solution.
Tell me about a time you had conflicting priorities?
During my internship as an ML engineer, the idea of the startup at the time was very amibitious and all the current research we could find in the field had less than ideal results, and we weren’t really able to replicate any of the research either. So this is just to get an idea of how daunting the project I felt at the time was given. It required a lot of effort, collaboration, and research before we could move forward with even a semblance of an MVP.
However, our CEO also very much liked the idea of the interns also participating on their YT channel as content creators. So we had free reign on the content and design, however we were expected to produce a video weekly. I’m not a content creator, so it was difficult to do all the prep work and make a video, narrate and edit. It would take almost a full day’s work to do the script, the content creation, editing, thumbnails etc. This basically meant that instead of 5 days we had 4 days to work on this daunting project every week. I communicated to our CEO that I understood he felt the content creation was important however it was taking a large chunk of time and energy away from our project. So I proposed that we cut our video time, did not provide a full script, and standardize the editing process and thumbnail creation. Resulting in videos now only taking half a day instead of a full day, this allowed us to make significant progress towards our sprints early on and get the MVP model, which we could then further train without sacrificing the content creation that our CEO wanted.
I really like the engineering culture at Meta, you get to work with some of the smartest individuals on huge projects that have significant impact globally. I also read the five core values and really resonated with them. In particular, be open, move fast, and focus on impact seem to be values I unconsiously try to align myself with already.