Patricia Cobelli, Developer in Montevideo, Montevideo Department, Uruguay
Patricia is available for hire
Hire Patricia

Patricia Cobelli

Verified Expert  in Engineering

Machine Learning Engineer and Artificial Intelligence Developer

Montevideo, Montevideo Department, Uruguay

Toptal member since November 8, 2022

Bio

Patricia is a self-taught machine learning engineer with expert knowledge of computer vision. She can handle every step of a machine learning solution, from data engineering to model training and deployment on AWS. Patricia is finishing her master's studies and collaborating on her thesis with researchers at Brown University.

Portfolio

Tired Banker
APIs, AWS IAM, AWS Lambda, Amazon Simple Queue Service (SQS)...
Pento
Python 3, PyTorch, Amazon Web Services (AWS), GitHub, JavaScript, Flask...
Universidad de la República
MATLAB, Python 3, PyTorch, GitLab, Machine Learning, Deep Learning...

Experience

Availability

Part-time

Preferred Environment

GitHub, Amazon Web Services (AWS), Python 3, PyTorch, Trello, Slack

The most amazing...

...solution I've worked on uses computer vision models to automatically enhance eCommerce product images, improving conversion and decreasing design costs.

Work Experience

Back-end Engineer

2023 - PRESENT
Tired Banker
  • Summarized documents using GPT-3, GPT-3.5, and GPT-4.
  • Developed the REST API and deployed it to AWS API Gateway with Serverless.
  • Created and populated the PostgreSQL database and deployed it to AWS RDS.
  • Developed the data ETL combining AWS SQS and AWS Lambdas, which were scheduled to extract data, transform it and store it in the database.
Technologies: APIs, AWS IAM, AWS Lambda, Amazon Simple Queue Service (SQS), Amazon API Gateway, Amazon RDS, FastAPI, SQL, SQLAlchemy, PostgreSQL, Scraping, Generative Pre-trained Transformers (GPT), OpenAI GPT-3 API, Natural Language Processing (NLP), Webhooks, Clean Code, REST APIs, React, Node.js, Next.js, Web Scraping, Website Data Scraping

Machine Learning Engineer

2020 - 2023
Pento
  • Planned and guided the development of an object detection library. It allowed data scientists with no computer vision knowledge to add valuable information to their pipelines without the overhead of understanding computer vision models.
  • Developed a web app that provides insights to improve online marketing KPIs. It involved data extraction and analysis, model implementation, API development with FastAPI, front-end development with React, and the deployment of all of them.
  • Created an open-source tool to track and compare embeddings. We intended the result to be something we would use, leading to a robust and easy-to-use tool.
  • Automated layer creation of an online mockup generator by implementing multiple computer vision models for segmentation, key point detection, etc. By developing the models, layer generation stopped being a bottleneck.
Technologies: Python 3, PyTorch, Amazon Web Services (AWS), GitHub, JavaScript, Flask, FastAPI, Neural Networks, PostgreSQL, React, Machine Learning, Deep Learning, Data Visualization, Data Science, Python, Artificial Intelligence (AI), Computer Vision, Back-end, Datasets, Data Wrangling, Pandas, Node.js, MySQL, Python Dataclasses, APIs, API Applications, Pydantic, Python Attrs, Image Processing, Machine Learning Operations (MLOps), Generative Models, Stable Diffusion, Open Source, PIP, Linux, PyCharm, Generative Adversarial Networks (GANs), Object Detection, Object Tracking, Amazon SageMaker, OAuth, SQLAlchemy, Image Recognition, Early-stage Startups, OAuth 2, Computer Vision Algorithms, Webhooks, Clean Code, REST APIs, Next.js, Web Scraping, Website Data Scraping

Researcher

2018 - 2023
Universidad de la República
  • Implemented physics-informed neural networks for wind field reconstruction from LiDAR data points, being the 1st person in my country to use this technology.
  • Applied Markov chain Monte Carlo algorithms (MCMC) to compute possible available power on a down-regulated wind farm, providing a single value output and a confidence range.
  • Taught Fluid Mechanics to a class of engineering students.
Technologies: MATLAB, Python 3, PyTorch, GitLab, Machine Learning, Deep Learning, Data Visualization, Python, Artificial Intelligence (AI), Physics Simulations, Datasets, Data Wrangling, Simulations, Pandas, PIP, Linux, PyCharm, Data Science, Energy

Data Scientist

2021 - 2022
Renovus
  • Developed endpoints for batch preprocessing of data, considering all the missing or incorrect data possibilities.
  • Guided machine learning models implementation. By combining machine learning and wind energy expertise, I could recommend a solution that proved to be better in a cost-effective way.
  • Developed an API that sends processed data to the front end.
Technologies: Python 3, FastAPI, PostgreSQL, Data Visualization, Data Science, Python, Back-end, Data Wrangling, Pandas, Amazon Web Services (AWS), MySQL, APIs, API Applications, Pydantic, Python Attrs, Machine Learning Operations (MLOps), PIP, SQLAlchemy, Early-stage Startups, Energy, REST APIs

Automatic Mockup Generation

A web app that allows users to mock up an image in various places. I helped develop models that automated the creation of Photoshop layers to create a mockup effect. This involved segmentation models, key point detection models, classic computer vision techniques, etc.

Open-source Software Tool

An open-source tool developed in Python that facilitates the tracking and comparison of embedding spaces. I participated in the development of a demo and a final code review, as well as handling new issues.

Object Detection Tool

I was the team leader for developing a PyTorch object detection library. We created a tool for data scientists to include object detection in their pipelines without computer vision knowledge required.

Web App for Digital Marketing Optimization

This project involved developing a React web app with a Python back end and PyTorch and Scikit-learn models. I was part of the small team that developed the app and was involved in every technical aspect of the product. Initially, I researched the data, studying relations between ad metrics and desired KPIs. Then, I created feature extraction pipelines and deployed them with Serverless. Finally, we developed the front end to show the data using React, Tailwind CSS, and Chart.js.

Computer Vision Model Hub

This project involved an internal model hub with different computer vision models used for sales purposes. These models included color detection, background removal, image classification, and object detection.

I developed and deployed the color detection model and image classification models. I also created the web pages where those models were used.

Wind Farm Analysis Tool

This tool involved a set of analytical models to predict the possible power a curtailed wind farm could produce. I was the tool's sole developer, creating models to predict the inflow wind speed, model the wake, and indicate the possible power on downstream wind turbines.

Renewable Energies Analysis Web App

https://www.renovus.tech/
This project involved developing a web app that centralized all the data related to wind and solar farms, from the maintenance documentation to the online metrics and weather forecasting. I created the wind energy back end with FastAPI.

Super-resolution Paper

I was the primary researcher for a paper assessing different generative adversarial networks (GANs) architecture for single image super-resolution. Two different GANs architectures, SRGAN and ESRGAN, were studied by comparing metrics based on results, inference, and training times.

Investments Analysis Web Page

This project consists of a web page with a complete analysis of S&P 500 companies. It includes stock prices, balance sheets, summaries from earnings call transcripts, and 10-Q files. I was in charge of the data pipeline, designing the database, scraping the sources, storing in the database, and creating the API to feed the data to the front-end.
2021 - 2023

Master's Degree in Energy Engineering

University of the Republic - Montevideo, Uruguay

2022 - 2022

Visiting Research Fellowship in Applied Mathematics

Brown University - Providence, Rhode Island, United States

2015 - 2020

Engineer's Degree in Mechanical Engineering

University of the Republic - Montevideo, Uruguay

2018 - 2018

Engineer's Degree in Chemical Engineering

Salamanca University - Salamanca, Spain

NOVEMBER 2020 - PRESENT

Using Python to Access Web Data

University of Michigan | via Coursera

NOVEMBER 2020 - PRESENT

Python Data Sctructures

University of Michigan | via Coursera

JANUARY 2015 - PRESENT

Certificate of Proficiency in English

University of Cambridge

Libraries/APIs

PyTorch, React, Pandas, Node.js, REST APIs, SQLAlchemy, Pydantic, Scikit-learn

Tools

Git, GitHub, GitLab, PyPI, MATLAB, Trello, Slack, PyCharm, Amazon SageMaker, AWS IAM, Amazon Simple Queue Service (SQS)

Languages

Python 3, Python, JavaScript, SQL

Storage

PostgreSQL, MySQL, Elasticsearch

Frameworks

Flask, Tailwind CSS, Next.js, OAuth 2

Paradigms

Clean Code, REST

Platforms

Amazon Web Services (AWS), Linux, AWS Lambda

Other

Neural Networks, FastAPI, Machine Learning, Deep Learning, Data Visualization, Data Science, Artificial Intelligence (AI), Computer Vision, Datasets, Image Processing, Back-end, Physics Simulations, Data Wrangling, Simulations, Python Dataclasses, APIs, API Applications, Python Attrs, Machine Learning Operations (MLOps), Object Detection, Image Recognition, Early-stage Startups, Energy, Computer Vision Algorithms, Webhooks, Web Scraping, Website Data Scraping, Generative Models, Stable Diffusion, Open Source, PIP, Serverless, Generative Adversarial Networks (GANs), Object Tracking, OAuth, AI Programming, Models, Natural Language Processing (NLP), Generative Pre-trained Transformer 3 (GPT-3), Generative Pre-trained Transformers (GPT), Amazon API Gateway, Amazon RDS, Scraping, OpenAI GPT-3 API

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring