Cesar is available for hire
Hire CesarCesar Romero
Verified Expert in Engineering
Artificial Intelligence (AI) Developer
Location
Seattle, WA, United States
Toptal Member Since
February 19, 2024
Cesar has over 13 years of experience in machine learning and software engineering. His career spans reputable industry giants like Amazon, Walmart, and Unity and smaller startup environments. Proficient in constructing large-scale classifiers and recommenders, Cesar has spearheaded initiatives involving the utilization of synthetic data for computer vision tasks. Currently freelancing, he prioritizes building maintainable systems that facilitate rapid iteration cycles.
Portfolio
R5
Python 3, GitHub, Unity, C#, PyTorch, Docker, Recommendation Systems, MLflow...
groundlight AI
Python 3, Kubernetes, Amazon Web Services (AWS), Amazon OpenSearch...
Unity
Python 3, Jupyter, Docker, Kubernetes, Amazon Web Services (AWS), Google Cloud...
Experience
Availability
Part-time
Preferred Environment
Linux, Emacs, Python 3, Docker, PyTorch, Kubernetes, Amazon Web Services (AWS), Jupyter, Pandas, Python, Data Analytics, Data Analysis
The most amazing...
...project I've led is the launch of the 1st ML service on AWS, including deep learning recommenders and computer vision systems with synthetic data.
Work Experience
Principal Machine Learning Engineer
2023 - PRESENT
R5
- Developed a tool to help detect near duplicate images in computer vision datasets, reducing the time to clean the dataset from days to minutes.
- Developed a tool to create datasets for different computer vision tasks and formats starting from video annotations, resulting in better benchmarks and more reproducible experiments.
- Implemented a new centralized dataset registry and process using DVC, S3, and GitLab, resulting in more visibility for management and more reproducible experiments for scientists.
- Designed and supervised the implementation of pipelines and processes for end-to-end offline experimentation, reducing the need for costly visual inspection and enabling data-informed decisions before new models were deployed to production.
- Created a plan to automate the process of producing and serving personalized recommendations, including A/B testing integration, reducing the frequency of updated recommendations from two weeks to daily.
- Created a plan to integrate novel AI behavior into a new VR game, enabling a personalized experience with an agent that can learn in real-time, interacting with the player.
Technologies: Python 3, GitHub, Unity, C#, PyTorch, Docker, Recommendation Systems, MLflow, Linux, Chief AI Officer, Prompt Engineering, LangChain, Pandas, OpenAI GPT-4 API, Convolutional Neural Networks (CNN), Predictive Analytics
Machine Learning Engineer
2021 - 2023
groundlight AI
- Worked on the project with the initial batch of full-time engineers, dealing with high ambiguity and adapting to multiple roles as needed.
- Built several aspects of the infrastructure, including monitoring and automated provisioning of edge devices.
- Redesigned and implemented critical parts of the core codebase to increase the productivity of scientists and enable more complex ML pipelines in production.
Technologies: Python 3, Kubernetes, Amazon Web Services (AWS), Amazon OpenSearch, NVIDIA Jetson, Docker, Fluentd, PyTorch, Django, Ansible, Robotics, Machine Learning, Data Science, Elasticsearch, Python, Predictive Modeling, Artificial Intelligence (AI), Data Scientist, Technical Leadership, Leadership, SQL, Generative Pre-trained Transformers (GPT), Machine Learning Operations (MLOps), Training, Large Language Models (LLMs), Fine-tuning, Applied Research, NLU, Natural Language Processing (NLP), Deep Learning, Neural Networks, Computer Vision, AI Model Training, PostgreSQL, Classification, Language Models, CI/CD Pipelines, Open Source, Open-source Software (OSS), Interviewing, GPU Computing, Git, Models, Object-oriented Programming (OOP), Research, AI Modeling, Search Engines, Linux, Convolutional Neural Networks (CNN), Predictive Analytics
Principal Machine Learning Engineer
2017 - 2021
Unity
- Worked across teams in the AI department and led collaborations with other departments.
- Spearheaded initiatives to direct ML and Unity developers toward creating tools for generating synthetic data to enhance computer vision using Unity.
- Played a role in refining the scope of domains and tasks for the initial release of the open-source perception package.
- Identified the necessity for a new internal simulation platform that supported various workloads, including distributed reinforcement learning (RL) and automated domain randomization. This led to the formation of a dedicated simulation platform team.
- Hosted weekly AI lightning talks and reading groups focusing on computer vision, Python, and software design, with monthly Ask Me Anything (AMA) sessions. These avenues aimed to enhance transparency and foster alignment within the AI department.
- Co-authored a proposal to fund a new company-wide ML platform team using tools like Kubeflow, Katib, and MLflow.
Technologies: Python 3, Jupyter, Docker, Kubernetes, Amazon Web Services (AWS), Google Cloud, Google AI Platform, Unity SDK, Computer Vision, Redis, Git, GitLab CI/CD, Google Stackdriver, Kubeflow, MLflow, Machine Learning, Public Speaking, AI Research, Interviewing, Data Science, Python, Predictive Modeling, Artificial Intelligence (AI), Data Scientist, Logistic Regression, Statistical Modeling, Technical Leadership, Leadership, Mentorship & Coaching, Data Manipulation, A/B Testing, TensorFlow, Google Cloud Platform (GCP), Machine Learning Operations (MLOps), Big Data, Training, Fine-tuning, Applied Research, NLU, Natural Language Processing (NLP), Deep Learning, Neural Networks, AI Model Training, Exploratory Data Analysis, eCommerce, Classification, Data Engineering, CI/CD Pipelines, Apache Airflow, Open Source, Open-source Software (OSS), Decentralization, Recommendation Systems, Text Classification, Hyperparameters, GPU Computing, Robotics, Image Generation, Models, Object-oriented Programming (OOP), Research, AI Modeling, Search Engines, Linux, Chief AI Officer, Pandas, Convolutional Neural Networks (CNN), Predictive Analytics
Software Development Engineer
2011 - 2017
Amazon.com
- Automated hyperparameter optimization for recommendation models.
- Used neural networks to produce personalized recommendations across categories and devices (Github.com/amznlabs/amazon-dsstne).
- Engaged as a member of the AWS machine learning service launch team. (Console.aws.amazon.com/machinelearning/).
- Played a role in the same-day delivery launch team (Amazon.com/sameday). Automated the process of determining export eligibility using machine learning.
Technologies: Java, Python 3, JavaScript, AngularJS, Amazon Web Services (AWS), Elasticsearch, Linux, Docker, Recommendation Systems, Text Classification, Hadoop, Machine Learning, Hyperparameters, GPU Computing, NVIDIA CUDA, Data Science, Predictive Modeling, Artificial Intelligence (AI), Data Scientist, Logistic Regression, Statistical Modeling, Data Analytics, Data Analysis, Technical Leadership, Leadership, Mentorship & Coaching, SQL, Data Manipulation, A/B Testing, Machine Learning Operations (MLOps), Big Data, Training, Fine-tuning, Applied Research, Natural Language Processing (NLP), Deep Learning, Neural Networks, AI Model Training, Exploratory Data Analysis, eCommerce, Classification, Data Engineering, CI/CD Pipelines, Open Source, Open-source Software (OSS), Decentralization, Interviewing, Git, Public Speaking, Models, Object-oriented Programming (OOP), AI Modeling, Search Engines, Web Development, Pandas, Predictive Analytics
Linux Administrator
2004 - 2007
Universidad Simon Bolivar
- Implemented a process to provision several new Linux workstations for students, resulting in a new lab of 16 computers all installed and configured in a single afternoon.
- Performed regular backups of student and professor accounts for the computer science department.
- Implemented a script to automate the backup of individual students and professors, resulting in more free storage on servers, which enabled higher quotas for all active users.
- Installed and configured a server to host internal forums to enhance the existing email lists, resulting in new ways for professors to communicate with students.
- Installed and configured a new DNS server for a new research lab, resulting in a new subdomain that could be used for professional emails and a new website to showcase the work of the researchers at the lab.
- Designed and implemented a new site for a new research lab, enabling the staff to maintain a list of projects that showcase the work done by the researchers.
Technologies: Linux, Python 2, Django, DNS, Subversion (SVN), MySQL
Experience
Video Stream Processors with Computer Vision
https://github.com/groundlight/stream/tree/mainI led the development that of an open-source project designed to process video streams and provide predictions, allowing human interaction without requiring any prior experience in computer vision. The project leverages a public API and is incredibly user-friendly, requiring a single command to launch a containerized application.
Unity Simulation
https://unity.com/products/unity-simulation-proI was a principal engineer deeply involved in the technical aspects of Unity Simulation from its inception to release. The idea was conceived in 2017, coinciding with my role as the inaugural engineer in the AI department. I prototyped the initial version of the simulation while drawing inspiration from academic research findings and recognizing its potential impact on computer vision research.
I spearheaded the research endeavor focused on training a cutting-edge object detection model using 90% synthetic data.
I spearheaded the research endeavor focused on training a cutting-edge object detection model using 90% synthetic data.
Synthetic Data for Computer Vision
https://blog.unity.com/engine-platform/use-unitys-perception-tools-to-generate-and-analyze-synthetic-data-at-scale-toI served as a principal engineer, leading a research and development initiative aimed at training a state-of-the-art object detection model using 90% synthetic data generated through Unity. The team comprised multiple engineers, a technical artist, a product manager, and a development manager, which led to the creation of three blog posts, an open-source project, and the publication of a paper. The publication was intended to share our efforts and findings within the computer vision research community.
Amazon Machine Learning
https://docs.aws.amazon.com/machine-learning/latest/dg/what-is-amazon-machine-learning.htmlI acted as the technical lead on the console of the first machine learning service on AWS, called Amazon Machine Learning. I led five developers working on different aspects of the web UX, including the onboarding experience, wizards, and data visualization.
After the launch, I assumed the position of tech lead to bring containers into the service to make it easier to leverage the ecosystem outside of Amazon. That was the beginning of what was launched two years later as SageMaker.
After the launch, I assumed the position of tech lead to bring containers into the service to make it easier to leverage the ecosystem outside of Amazon. That was the beginning of what was launched two years later as SageMaker.
Education
2008 - 2010
Master's Degree in Computer Science
UCLA - Los Angeles, CA, USA
Skills
Libraries/APIs
PyTorch, Pandas, TensorFlow, OpenCV, D3.js
Tools
Git, Amazon OpenSearch, NVIDIA Jetson, Fluentd, Emacs, Jupyter, Ansible, Google AI Platform, Unity SDK, GitLab CI/CD, Google Stackdriver, Apache Airflow, GitHub, Subversion (SVN)
Languages
Python 3, Python, Java, JavaScript, SQL, C#, Python 2
Paradigms
Data Science, Object-oriented Programming (OOP)
Platforms
Linux, Docker, Amazon Web Services (AWS), Kubernetes, YouTube, Kubeflow, NVIDIA CUDA, Google Cloud Platform (GCP)
Storage
Google Cloud, Elasticsearch, Redis, PostgreSQL, MySQL
Frameworks
Django, AngularJS, Hadoop, Bootstrap, Unity
Other
Machine Learning, Artificial Intelligence (AI), Computer Vision, Interviewing, Text Classification, Predictive Modeling, Data Scientist, Logistic Regression, Statistical Modeling, Data Analytics, Data Analysis, A/B Testing, Training, Fine-tuning, Deep Learning, Neural Networks, AI Model Training, Exploratory Data Analysis, Classification, Models, Predictive Analytics, Algorithms, MLflow, Public Speaking, Recommendation Systems, Hyperparameters, Technical Leadership, Leadership, Mentorship & Coaching, Generative Pre-trained Transformers (GPT), Machine Learning Operations (MLOps), Large Language Models (LLMs), Applied Research, NLU, Natural Language Processing (NLP), eCommerce, CI/CD Pipelines, Open Source, Open-source Software (OSS), Decentralization, Research, AI Modeling, Chief AI Officer, Prompt Engineering, LangChain, OpenAI GPT-4 API, Convolutional Neural Networks (CNN), Optimization, RTSP, Robotics, AI Research, GPU Computing, Data Manipulation, Big Data, Causal Inference, Language Models, Data Engineering, Image Generation, State Machines, Search Engines, DNS, Web Development
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