elevator pitch Flashcards
Can you walk me through your Bayesian Regression model and how you integrated ElasticNet regression?
Well, first off that’s a great question. I’ll talk about the elastic approach first the elastic net is a combination between Lasso and regression. The point of the elastic net is to work with highly correlated data. A lot of our data is of course face data so your nose is not much larger than your eye your jaw is not much larger than your forehead so a lot of that data exist within the same scope or length and that’s why we chose the elastic net model because it was affected when it came to looking at those me and seeing how they correlated with certain features.
The vision approach is the opposite of what we call a frequent test approach for frequency approach to tell you so you were looking at housing data you put that through new Network can you tell me what your prediction is for how much this hassle cost and it’ll do it things and they’ll tell you hey this house should be about 400 K. I have a 94% accuracy that is between 400 K etc. and the basement approach would look at it and twell this house has a 70% chance of be being 400 K but there’s also a 15% chance that it could be less than 400 K and it could be a 15% chance it could be more than 400 K, the goal of that is to see the sort of risk involved because as we know not everything is perfect. A lot of the data that we are using is actual data but the ratings or how people perceive a certain face or facial feature is subjective there’s a lot of cultural biases that come into that so there’s not a 100% hey this person is very handsome or this person is not attractive, etc
How did you achieve 95% accuracy in your facial analysis pipeline? What challenges did you face?
oh yeah honestly getting started was really the hardest part because there’s a lot of software out there that’s good for skin segmentation but it might not be the best for the accuracy that we’re looking for a lot of software can do it, but can they do it right are they missing chunks out of their face when you know when you run this image through their software so finding that software that could support our research was the biggest hassle
we ended up deciding on the library called sKone, SKone is a skin segmentation algorithm, but what sets it apart is that skone uses LPBH and harr cascade based segmentation. Not only or using these methods accurate but it also allows for the product or the output to be in RGB format. After that, we convert the RGB to C i.e. lab color space. This is a color space that represents human skin the most, and it’s overall the best sort of metric for us in this process.
How did you use Python, Pandas, and SQL to automate quality control processes? Can you describe an example?
Yeah, absolutely using Python and sequel was a big part of my role at my tech during my time there I was tasked with a couple of things, but one of them was to sort out the data for next year European immigration so that’s when my role sort of let me in the direction of working with those cross functional teams and really understanding the habits that they see under data whether it’s human error or any sort of irregularity under data and also what sort of forms or technology need to be improved in order for a better ecosystem for our data base.
I use Python in conjunction with QD which is their ERP system QAD is known for his flexibility. It’s quite old. I would say so. It definitely gives wiggle room for a human error and certain aspects but not in other so a lot of my time in Python dealt with looking through a human errors and trying to find a consistent basis for records, as well as finding lost records along the way. This was in part due to like a QAD certain times of people would know how to bypass certain restrictions like how long an address can be and often times that data is just kept with the individual so when they left that information left with them. So you would have missed maths records when he came to accounting or reporting often times a lot of the amounts that were being generated in the reports were not to be believed or credible, just cause we sort of realized that hate this CRMQD feed, salesforce, which feeds our data warehouse and so much other more that when that sources corrupted it corrupt the whole ecosystem there.
What was your role in leveraging Salesforce and QAD to enhance efficiency?
yeah actually we are QAD and salesforce are a sort of tie together. A QAD is our ERP system at my tag and that feeds salesforce which feeds a lot of of the other data hubs within the ecosystem.. QD records are sort of a little bit difficult to work with if you wanted to change a record by itself you could do that but mass changes are very difficult and with the system being as old as it is, nobody really knows how to do that within QAD itself and so we had to use a different software that can make those changes in QAD but only certain people had permission to die first of all then also, that’s Alfred would only be run overnight. Meaning that if a report were needed, I could not be generated because like I said making mass changes would take a very long time and it was very computationally expensive. That’s why it was best to run overnight.. so what I did was that I sort of single that the accounts that needed changes being made a Python script to look at the common human errors that we can see by for example Boulevard some people were spelled all the way out and some people would just B like BLVD and sometimes QD would not pick up on that when we needed to run reports so often times those reports were inaccurate.
How did you create a standardized swim lane for Next Gen ERP integration, and how did it improve workflow consistency?
yeah absolutely the base of the swim lane was with how we first entered a customer into our system. We had these first day on paper. I think they still on paper and then some you could do it on Excel, but both of them are not that good? We have these sort of forms that people fill out whenever there’s a no customer being added to our system. The problem with that is first of all the form is not very restrictive and secondly, the form is not very descriptive mean that a lot of the necessary information needed about the customer is not really asked on the form, the forms are also sort of leading to the problem because they are updated in many ways new system sort of requires certain nomenclature when it comes to the integration process of customer data and the forms are not really efficient when it comes to making making it out of the person or editor’s hands and effectively conveyed that data to our data warehouse or even our QAd systems.
Can you explain how you used Selenium and Pandas to automate content extraction for web development projects?
How did your web scraper improve high school databases? What was the impact?
Can you describe the database and login system you built for Connect? How did you ensure security and scalability?
How did you secure sponsorships and partnerships for the football club?
What were some challenges you faced while coordinating practice sessions and tournaments?
How did you establish a collaborative lab environment for AI research?
What is your research on AI-driven facial analysis and perceptual biases aiming to solve?
You list Power BI and Tableau as skills. Can you share an example of a project where you used these tools?
Yeah, funny enough. I don’t think a lot of people share this experience but the first time I use sort of these visual visualization tools was with tableau at the time I had a class called quantitative method for business. I was about freshman so I was using Excel quite a lot just for analytics and just learning about business statistics, and a part of that was to incorporate tableau with Excel to sort of visualize our data at the time. I never knew that Pablo was something that I would ever use again so coming back full circle is pretty cool.
Overall, since then I would say that I have not used tableau a lot. I think power BI sort of might go to just because it connects with Asher and cloud in so many tools within the Microsoft space that when I was working at my tech, it just felt align to my tax goals and missions and overall RBI was amused with a integration to their data warehouse. RBI was sort of the reporting tool that would be one of the final steps within the process of generating reports so a lot of the information that we used on the RBI dashboard was fed by QAD and salesforce
How comfortable are you with SQL? Can you describe a time when you optimized a query for performance?
Yeah, absolutely working over the summer at my tech really made me comfortable with sequel because like I said there was quite a bit of data that I had to look at and sequel and just PBI sorting through that information was definitely a hassle sequel was a great resource I would say, but once again, Python was really what I leaned towards because I think Python has a little bit more flexibility when it comes to using different library in different methods to finding what you need.
Can you describe a time when you had to troubleshoot a technical issue in one of your projects?
yeah absolutely when I’m working on my first or my current project with computer vision I call a face value that project in the itself was like so foreign to me because a lot of times you’ll see like there’s creators online and be like hey I built a calculator. Hey I did this. I did. I built the pipeline that leverages post credit post grass, etc. if you like OK if that’s tangible a lot of people havI could probably go to get hub or any sort of stack overflow type website and look into the app be like oh so this is how they did this. But when I first started this project, it was just me and my professor and at the time and even currently there is little to no sort of documentation of the work that we’ve done so a lot of times I would have to manually check everything check that the results produced by our calculations or even just our measurements are in the right sort of matrix and this would be very timely, but come to find out that a lot of the subject or even the sort of paper is written about what we’re doing are scientific papers. A lot of that data is submitted to whoever is the authority for distributing that paper and nobody else knows it so it kind of feels like you’re walking on uncharted ground and I’m just so thankful for my professor being a part of a team that has worked on his project before his knowledge you know how has a allowed me to kind of extend this project into a new space that it hasn’t been done before.
Have you worked in cross-functional teams before? How did you navigate collaboration challenges?
yeah I have worked with a cross functional teams during my time at my tech my tech is a global company so we had branches in North America, Canada, AAP, Europe, Africa, etc.. while it did sort of slow down communication a bit as you know not everybody is on the same time. I vividly remember instances where I was trying to communicate with the Mitech members in Asia, Pacific or even in Europe and a lot of times sending out those emails or communications. I would know that hey I’m probably gonna get this when I clock in tomorrow morning that’s when I see that message so just being able to work with them was in itself a great learning experience because it allows you to experience new cultures I vividly remember just meeting with the people from my tech Vietnam and even though there was sort of a language with their just how they’re supposed to their job was unique in a breath of fresh air to me. Also when working with those international teams a lot of their roles when it comes to just leveraging technology and my tech and distributing that technology, they built a lot of great knowledge over the years so being able to talk to them actually I learned a couple things to buy why I certain things are the way they are at my tech North America, which is surprising because like I said before we’re trying to change that, but my tech is sort of the company that has a lot of tribal knowledge and sort of there’s a sort of chief information about this particular subject. When a person leaves you know a lot of acknowledge is just stuck with them and they go on being able to sort of understand their point of view as a benefit.
What kind of roles are you looking for after graduation?
I am looking forward to data analyst roles. I believe these roles is suited for me because not only am I engrossed in new technologies and methods every day, but also the job helped me communicate with people and also give them the sort of information that I hold I think it’s really important to effectively communicate with a individual whether just explain something to them or even discover discussing new topics effective communication is something that we as humans prideso I think the job in itself a house with that. Also, with just a technology aspect, I envision myself working as a data analyst, but also besides my job I also want to use the technology that is presented to me to build projects to make a difference in just a tech equal space but also just the world in general I think data analytics is getting to the point where the AI missing learning is being incorporated into it, and that is room for investment in the field.
What excites you the most about data science and machine learning?
What does size mean most about data science and machine learning is that it’s applicable to a lot of different spaces within society as well as within just human authorization as well. I’ve seen a lot of people get creative with what they will develop with machine learning, and overall, I think that the investment of machine learning allows students to build tools that can compete with industry level talents with the knowledge that they gain in school.
Also, another reason why I like machine learning is that it can be used for various purposes just with my purpose currently I’m using it for a social and economic awareness. I think it can be used in many different fuels such as medicine but just helping out disabled people or even discovery, new cure diseases machine learning is something that is going to be at the forefront of that.
