Welcome To My Portfolio!

Name: Oscar Casas

Profile: Data Scientist

Email: casaso@berkeley.edu

Phone: (970) 708-4547

Skills

Python 95%
R 80%
SQL 100%
Cloud Computing 75%
ML & AI 75%
Tableau 85%
About me

Professional: I am a full-time Masters of Data Science graduate student at UC Berkeley and I am a full time data analyst at Lockheed Martin. (How do I balance the two? idk but sometimes crying helps) Just kidding! I love my workload it seems like everyday I learn something new.

Personal: When I am not working, I love everything there is to do with the outdoors. Whether it is freediving, piloting or skiing, I love to fill my freetime with activities I can enjoy with friends and family.

Mix: Although I like to leave myself enough time to enjoy my weekend activities, right now I am focusing on a facial recognition system that identifies refugees in camps so family members across the world can use old photos to identify them. If the project leads to just one reconnection it will have all been worth it. The program already works, my work now is just talking to non profits to help me implement it.

Specialties

The subjects and areas I feel comfortable working in.

Machine Learning

From simple regression or bayesian networks on one target interval, to complex multitiered nueral networks, I have worked on different projects have learnt a lot along the way. Whether it is using ensembles on segmented data of different sizes to reduce misclassifications to performing text mining on repair notes to then predict part failures my background is a bit diverse.

Big Data

With more data, hard drives can hardly handle our projects. I have learnt to build pipelines using both GCP and AWS. It is important to be able to run large quantities of data and processes off hard drives so I have become comfortable using docker, hadoop, spark, lambda and dynamo db to name a few of the different levels of architecture I like to use in my data pipelines.

Model Deployment

It's one thing to build a cool model, then it is completely different to bring wide range usability to it. I helped create an API capable of using AWS servers where teams can upload in-house processes so others can access them in the API and readily use and share them. I have also created my own websites that implement facial recognition models using Django.

Data Wrangling & Analysis

Through my experiences there has always been one thing that seems to always be true. Data received is never clean. I have learnt just how important it is to analyze and clean the data before beginning a model. It's not just finding outliers and missing values, its going the extra mile and trying to find if the data collection was correct, finding errors in previous assumptions and dealing with them.

Data Visualization

Building models may be the bulk of the work, but presentation is the bulk of the effect it may have on stakeholders. I have learnt to use tableau, matplotlib, seaborn and ggplot to represent the models I like. While static visualization is usually enough, I have also focused some effort on learning dynamic visualizations to bring an extra layer into my presentations.

Statistics

Statistics is a broad subject, but it is a vital part of my analysis throughout the whole model building process. R is my favorite tool for statistical exploration to make sure the assumptions I have for a model fit the specifications of an algorithm. I also prefer to use R for statitical analysis to truly understand the results of my data. While I prefer to use R, I feel comfortable in Python as well.

WORKS COMPLETED

YEARS OF EXPERIENCE

PROJECTS LEAD

AWARD & RECOGNITIONS

Projects

A couple of the fun projects I have completed. Most of these projects are short just because the majority of my work/projects is at my job and the security issues wont let me showcase them but I am more than happy to talk a bit of the projects I have done (to the extent I am allowed to).

FamUnite

Facial Recognition / Jan 2021

PACMAN DQL

Machine Learning / Dec 2021

Title Analysis

Stat Analysis / Dec. 2021

GCP App Pipeline

Cloud Comp. / Dec 2018

Connect 4 AI

Minimax Algorithms / June 2021

Twitters Effect on Dogecoin

LSTM/NLP/ML / Sep 2021
Daniel Garay

Oscar has been a key component in our teams project in developing our e-gaming crypto, bitgames, and understanding the statistics behind our project. Oscar is always the first one to find potential problems and offer solutions. He continuously impresses the team by how quickly he learns concepts and is able to apply them to our development process. His work in developing the AI for bitgames has been pivotal and he has helped the team be ahead of the cusp. We hope to keep working with you and looking forward to see you grow more in our team. Thanks Oscar for your wonderful work!

Lockheed Principal Data Scientist

I have been super impressed over the past few months with the level of initiative Oscar has shown in several development efforts. Oscar has been driving innovation in ---- with our API interface to ------, within the ----- Augmentation effort where he has spearheaded our Text Mining and ML build effort, and within the ---- effort where he has provided significant contributions to overall team success. Oscar has been a model of what it looks like to buy in to our desired culture and raise the bar in every development effort he has been a part of. Your work has put our team in a position to achieve required project results on numerous fronts and is helping us achieve our Key Results. Thank you for Reshaping our Operations and keep up the great work!

Lockheed Data Analyst Manager

I would like to take this opportunity to commend Oscar Casa for his continued performance on the ----- Team. Oscar came into the ----- team and immediately hit the ground running making a splash in everything he has taken on. Oscar immediately took ownership of several projects working with LM Consultants and providing subject matter expertise and has continued to lead the charge on bringing that capability to fruition. Thank you, Oscar, for everything that you do!

Blog

A couple of blogs I've written this year regarding my experiences working from home and learning new skills.

Worklife

Experience of Starting a New Job Remotely

Covid forced many people to a work from home environment, but for some going from college grad to a full time remote worker can be difficult and confusing.

Machine Learning

Too Many Nominal Targets Not Enough Data.

If you find yourself trying to predict a nominal value on many potential options, multi-tiered segmentation might just be the solution you're looking for.

Worklife

Balancing Masters Degree and Work

Balancing a masters degree and a full time job can many times feel overwhelming, but learning to not just balance but leverage the side by side learning to improve in both aspects can truly be the edge that can not only save time but improve performance.