Behavioral Interview Flashcards
Tell me about yourself.
So I graduated back in 2016 from UVA where I studied Econ and math. About six months after I graduated I ended up taking a job down in North Carolina as a data analyst and this is a really incredible experience. It’s what got me into the world of business intelligence, data analytics, and operations research.
After a while, though, I felt that the next step in my career would be to get into data science which was getting pretty hot at the time. So I went back to UVA got my masters in data science and then since done worked as a machine learning engineer with Booz Allen.
Fast-forward a couple years and I feel that now is a good time for me to get back in the job market. There’s a few different reasons for that and we can talk about that in a second but ideally I’m looking for a role where I can get back into data analytics because I’ve sort of learned that that’s what I’m most passionate about. But then also potentially doing some modeling or predictive analytics work work at leverage my more recent ML engineering experience. So that’s what brings me here.
What are your strengths and weaknesses?
Strengths:
I think my strength for this particular position is my work experience. On one hand, I have several years of experience with traditional data analytics. And that’s valuable its own way, but on top of that, I think there’s a lot of synergy between the work I’ve had as an ML engineer on this particular position. Maybe you’re doing and doing some modeling on your own or at some point you might be working with other data scientist and engineers and being able to understand their language and communicate would be super important.
Weaknesses:
Yea, I could name a couple things. Firstly, I’m not the best at presenting. I can do it, but it does take me some time to prep an advance to be able to do well and if you would’ve maybe put me in a position where last minute I got a present on something that I’m not familiar with I might have some issues.
Secondly, sometimes i tend to overcomplicated things.
Name one time you’ve failed
So this was a little early on in my time in North Carolina remembered one point our directive operations will present on what he needs to during quarterly board meetings. He ended up texting me and he said he needs some stats ASAP. Long story short I really wasn’t able to answer it during the time in the board meeting. It was something that I should’ve had already. So remember for the next meeting I put together Won Seagal filed that contained at least six or seven queries all ready to go so that if you ever ask me a question, the query already be set up to do is change whatever parameters.
Name a conflict on the job
Well, let me just say the structure of the agency down in North Carolina was very interesting and it did lead to some friction. So the number of people that actually work for the agency directly probably less than a dozen. Most if not all of the office staff or consultants, which makes things a little interesting. Only because office politics are always some people that wanna get ahead that’s part of it. It’s even more pronounced. We have a bunch of different consultants from different firms because at the end of the day not only do you wanna personally get ahead, but doing well means you’re consulting company gets more work.
And I remember one of the initiatives that we had was dashboard development. Through this process was teamed up with another consultant to basically bring this thing to life. and I remember at the time that this consultant wasn’t really the most technically gifted, which is fine but what I started to realize was that overtime the way to put simply this consultant with maneuvering trying to set up updates with the client sometimes without me, knowing and then also putting particular emphasis on leading and presenting update meetings. They ended up being a mismatch between the work and the effort that was actually being put forth in this project and then just trying to get in front of the client and showing what was actually done.
Describe BAH Role
While I was at Booz Allen, nearly all of my time was spent working with the Department of Veteran Affairs. The primary project that I was involved with had to do with patient cost modeling. So in essence, using historical patient related data to predict cost of care over the next 6-12 months. The model development process is obviously a part of the work, but an even bigger part of the role was actually data engineering. So taking in different datasets that was made available and transforming it into a state where we can train models with. On top of the ML modeling and the data engineering. There’s also some miscellaneous data analytics and reporting task that I help out with, but that was obviously much smaller part of my role.
NASA Capstone
This project was part of our Capstone during my masters. The overlying goal was the improve estimates of PM2.5, which is essentially refers to fine inhalable air particles.
Currently the way that PM2.5 is measured is through ground sensors throughout the country. These sensors are very accurate but obviously expensive and not very scalable. Now basically a bunch of very smart PhDs learned that they could accurately retrieve aerosol estimates including PM2.5 using polarimeter data. There’s something called the PACE-MAPP algorithm that allows you to do this. But the problem is this model is very computationally intensive and so the idea of our project is to replace the inefficient portion of this algorithm with a neural network.
And we can go into more detail here in a bit but long story short we were able to speed up the heavy portion of the algorithm by over 1000x while maintaining a very high level of accuracy
Why Capital One?
Well, I do have a few different filters that come to mine when it comes to job hunting. First and foremost is geography and I’ve grown up in manassas, Virginia, and I’ve been in nova most of my life and this is the area where I want to stick in for the long-term. Then the next question is the type of company. And while I’ve been in consulting for most of my career, when I stop to think about what’s best for me in the long-term for career progression in tech, it’s important to stay in the cutting edge. And from what I’ve seen so far, generally speaking, the most technically rigorous companies are private tech companies.
So that leaves a few different options you have Google you have a small Microsoft presence, Amazon is always there. And then you have Capital One, which is basically a tech company in my eyes.
And then there’s culture and work life balance. And I have several friends that work at that one my brother works there. I’ve heard nothing but good things.
Why leaving current role?
Well, with any job, you have the good and the bad. But one of the things that I’ve had issues with as a machine learning engineer is that you become so removed at least in my experience from the actual business problem and you’re purely focused on optimizing and solving the engineering problem. You sort of lose the why in what you’re doing.
And so while I do still like the modeling and the coding, ideally I’d love to have a position where I’m a little bit more in the middle when it comes to working on the business aspect as well as the engineering aspect.
Now I am looking to move from working for the public sector towards a private sector that has more to do with the development process. I’ve learned from my own experience and then also talking to a lot of my friends who are in consulting, the development process can be so brutal
What do you bring to finance?
First off while I don’t necessarily have a specifically finance experience, I’d say the software and technology that I’ve used in my experience as a data analyst will translate very well into a data analyst in finance or banking. Everything from Excel tableau SQL Python, Spark, ML they all work well across data analytics in any domain or field. And I am very substantial, professional and educational experience using all of these tools.
At the same time, there is a certain type of thinking, like a thought process that starts to develop after you’ve been doing data analytics for so long. That’s something that can’t really be taught. It really just comes with experience and it’s built over time. That also applies well to any different field in which you’re trying to apply data analytics or data modeling not just the specific domains that I’ve worked in.
And admittedly, while healthcare is pretty different from finance, but I will say the total experience that I have beyond the surface is a little bit like finance in certain ways. Apart from the tolling statistics and roadway, statistics that we put together a big portion of the data insides that we look into are revenue and financials. Things like what’s the expected amount to be collected from our transactions over the span of six months 12 months 18 months. We send out invoice invoices to postpaid customers until we look at the percent that’s paid after 30 6090 days. We look at the fees that accumulate from late payments. And this isn’t that much different than consumer banking where some of the metrics you’re looking at are know the balances of credit cards interest that accumulates late payments. There’s a lot of synergy there and I think that the experience that I have will apply pretty well in consumer banking space.
Difficulty with BAH development cycle
One of the things that are found particularly brutal when working at Buza with a development cycle for our ML project.
So just to give you some context, we were initially developing within a Palintir managed environment. And this started out all right because they had several very enriched curated data sets that were available that we could use for our training. So this initially made the data engineering aspect, pretty straightforward, and we were quickly able to start working on the actual model building. Unfortunately, what we realized was to put it quite simply the environment that was set up for us wasn’t capable of reliably building and rolling out ML models. So we went on for some time trying to make it work, but eventually we got to the point and there were some other teams that also had these issues so there was a decision made to migrate to a different platform. Where our data would be stored in some Microsoft Azure storage and we will be doing our pipelining and model building with databricks on top.
And while in the long-term, this was gonna be better off for everyone the short term transition was very messy. If you know anything about the VA you’ll know that they’re very heavily silo and the technology portions is no different. And so trying to migrate the data that we needed was a significant issue on top of having to rebuild some of these curated data sets that were initially made out for us. And so for a while, we were put on pause because we just couldn’t have the data and I ended up having to work on some other projects in the meantime, but eventually got to the point where we sort of felt like we had to do something because we didn’t want to risk our project just getting lost in the transition process.
So one thing we recognized was that the curated data sets that we initially were working with had a lot of columns that just never ended up being used and so what we ended up doing was a sort of brutal process, but we were we started to analyze the code and then begin to strip out The columns that weren’t necessary for our training. And by doing that we got to the point where there is only about eight or nine different data sets that comprise the columns that we needed as opposed to almost 40 or 50 different data sets that made up the entire curated table.
And so we were able to go to the VA and say look we just need these six or seven data sets to get started up with and we could start doing our work and then the rest could follow, but just please prioritize this and so we were able to do that.
Example of conducting research
The North Carolina Turnpike Authority has both operations and IT vendors that support them. And there are times where this result in some conflict between the two when it comes to enforcing their contractual obligations. So for example, one of the processes that our operations center is tasked with is image review. So when a vehicle drives through a toll road, and the automated processes aren’t able to identify the vehicle on the license plate that image is sent to a person and they look at the image and actually have to manually type in the license plate and then submit it. The people who do image review have to meet certain requirements on speed and accuracy. But then you have our IT vendors that actually provide the IT equipment and computers for the image of viewers to do their work.
And at one point there were some technical issues that were going on at the time and so on one hand, you had the operations managers saying that hey we can’t meet our speed requirements because the actual software and equipment is just too slow. There’s a lot of lagging and it’s slowing down our work. Then on the other hand, you have the actual IT vendor saying look this stuff shouldn’t be impacting the software to this degree and it’s only certain times in the day where we’re having these outages.
And so the agency was tasked with me to do some research and see whether I could figure out whether it’s truly an operations or an IT issue.
So basically what I did was look at image review statistics over a larger period of time. In specifically, I looked at a viewers that had high speeds and low standard deviations of their speeds over several weeks and months and then I compared this to the time period. When we were supposed to be experiencing those technical issues and it turns out that there was a pretty significant drop in the productivity of all of these image reviewers across the board.
Challenge At Work
So probably after say about a couple of years of working in transportation and tolling I was finally at a place where I felt like I understood the data very well and I understood the systems in which the data flow through well.
But then one of the challenges I had was trying to roll out some of this knowledge to other departments in the agency. So for example, when you ask the question, how many transactions did we get in the past quarter that answer could vary and it really depends on the context in which you’re asking. In a story to realize different departments were using reports to report numbers that sometimes weren’t appropriate and other times were completely wrong. and it’s hard as someone that still so young to come up to someone that’s been reporting the same numbers using the same report for 20 years and tell them this isn’t right.
But I sort of realized sometimes it’s on the in the approach and how you say things versus what you actually say. So for the particular groups that I knew weren’t necessarily using this the right number instead of just coming up and saying straight out pay, this is wrong you need to correct it. You certainly got a phrase it a little bit differently so in this case, I basically said look, this is actually the way in which are Systems are set up and I sort of gave like a high level description of it and then I sort of pointed to where respective person was getting their numbers and I just sort of just phrase it as this would be an even better measure of what you’re trying to report and the reasons are XYZ. And for the most poor people were on board it was only about one or two very senior people that continue to report as is and at that point, we just press forward in our in our reporting because quite frankly, we knew what we were doing was correct and we had the substance behind it. At this point, we have let everyone know the reasons behind our numbers in our reporting by getting in front of people and just giving a brief overview of the transaction systems and then we let any contradictions with other numbers just come up organically which they did and at that point we had a pretty very high-level discussion with the executive board on How we were going to report and moving forward.
Who are you going to be in 5 years?
Career to be in 5 years?
Need to fill out.
Describe a time when a team member was struggling.
Work on this
