Carolina Arriaga

Data Scientist | Vallejo, CA | caro.arriaga@gmail.com | github:caroarriaga

As a published researcher, former entrepreneur, and serial hackathon winner, I consistently deliver exceptional results in diverse stakeholder environments. I am eager to join a team to facilitate data-driven decisions, contribute to impactful outcomes, and expand my expertise.

Skills

Python:
  • Spark
  • Pytorch
  • Streamlit
  • Pandas
  • Bokeh
    • Other languages:
  • R
  • SQL
  • BigQuery
  • HTML/CSS

    • Experience

    –

    Independent researcher at Progressive Emergence

    Highlights

    • Lead author of a 2023 working paper on electrification of an early-adopter fleet in Canada.
    • Co-authored a paper for IEEE EDUCON 2020 on evaluating quality decision-making and applied sentiment analysis.

    –

    Co-Founder, Data Scientist at Facets

    Highlights

    • Implemented data analytics to web, and react native apps using Segment and Mixpanel.
    • Conducted product fit surveys, interviews, and data analysis leading to major pivot decisions.

    –

    Data Scientist, Zero Emission Vehicles Intern at The ICCT

    Highlights

    • Designed and implemented an energy rates database covering all Canadian provinces.
    • Deployed a web app with interactive visualizations to help fleet managers assess EV savings.

    –

    Senior Integration Engineer, Cooling at Whirlpool's Center of Innovation and Technology (CETEC)

    Highlights

    • Designed and ran causal experiments using regressor models. Saved 9m in cost reductions.
    • Led the simulation optimization to comply with 2020 DOE energy and refrigerant standards.
    • Data mined refrigerator reviews and ratings and built a rating predictor for new products.

    Education

    –

    Master in Information and Data Science from University of California, Berkeley

    Courses

    • Leopard Spotting: A Deep Learning model to re-id leopards. Used roboflow, YOLOv5, combined with ResNet-18 CNN embedding model for re-id. We achieved a top-1 0.61 accuracy, on par with SOTA on other species.
    • ShapSum: A framework to predict human judgement multi-dimensional quality scores for text summarization: Used a random forests model coefficients and shapley value analysis to understand what NLP metrics evaluate a summary so its on par with human judgement.
    • A study on people's memory after exposed to a controlled tweet: we found that people tend to remember false information with positive sentiment.
    • Effectiveness of government interventions and Covid-19 deaths: Used a linear regression model, we found that in early pandmic, states with interventions saved more lives than states that didn't had them in effect.

    –

    B.Sc. in Engineering Physics from Tec de Monterrey

    Courses

    Awards

    Winner - Honorable mention from DataPalooza

    Winner - 1st Place from AI and Identity - Okta Hackathon

    Winner - TinyMCE Challenge and Clarifai Challenge from API World + AI DevWorld Hackathon

    © 2019. All rights reserved.