I am a Machine Learning Research Scientist and Statistician, currently working as a ML engineer for image processing in the automotive industry.

My background is in pure Statistics and Data Science. My areas of interest in ML are related to R&D of new optimization/learning strategies in domains such as computer vision, time series prediction, NLP and advanced predictive modeling.

I have had the opportunity to work on different machine learning problems in industry and research, where my main focus has been developing and productionalizing deep learning models, specially in computer vision.

If you are interested in doing research with me. In ML I am interested different paradigms such as transfer learning, multi task models, deep adversarial robustness, generative models, deep neural networks compression and continual learning. In statistics, I am currently researching methods and applications related to time series prediction, Bayesian inference, continuous dependence modeling (copula theory) and sampling strategies.

However, I am open to all types of problems that involve finding and analyzing complex patterns hidden in data.

Experience

Machine Learning engineer

Asaphus Vision GmbH, Berlin, Germany (12.2018-Present)

R&D Deep learning
Machine learning Operations (MLOps)
Implement end-to-end ML Convolutional Neural Networks (CNNs) pipelines for facial expression recognition, seatbelt detection, eye Blinks detection (Semantic segmentation)
Model compression and efficient inference (model optimization, sparse and quantized deep neural networks, statistics to determine quantization loss in precision)
Improving and maintaining ML pipelines (ex. Tensorflow pipelines for facial recognition and face detection, creating tools for deep learning framework exchange formats, TF custom layers, etc.)
Neural preprocessing and augmentation: Image translation (Generative Adversarial Networks and Neural Style Transfer) to simulate lighting conditions, transform image styles (RGB to Near-Infrared NIR)
Machine learning Interpretability - Explainable AI

Research Assistant Scientist

Universidad Santo Tomás, Bogota, Colombia (07.2015-12.2018)

Research in statistics (Main areas Time series prediction and Bayesian inference)
Bayesian continuous dependence models and Copula theory applied to time series prediction and risk management
Bayesian estimation for Jump-Diffusion processes
Natural Language Processing in time series prediction

ML/Statistics advisor and Freelancer

Independent and free lancing projects, Bogota, Colombia (10.2015-Present)

ML advisor in B.Sc Statistics thesis: Deep learning attention mechanisms and interpretability for retina damage detection in Optical Coherence Tomographies (OCT)
Multivariate Time series prediction and Bayesian statistics (epidemiological data)
Statistics Faculty Assistant
Statistical advisor for PhD theses in social sciences and administration

Education

M.Sc. Data Science

Universität Potsdam, Potsdam, Germany

I focused my work and research on using Deep Learning to solve different problems in image classification/generation, reinforcement learning, time series prediction with NLP and signal processing. I have spent most of my time working on developing and studying methods in multi-task learning, GANs, catastrophic forgetting in deep models, transfer learning, adversarial examples and defenses and Machine learning interpretability.

B.Sc Statistics

Universidad Santo Tomás, Bogota, Colombia

I centered my research interest on stochastic processes, Bayesian statistics, continuous dependence models (Copulas), multivariate time series analysis and sampling strategies. I also learned traditional Machine Learning methods and analytics.

Strengths and Interests

Deep Learning and classical machine learning

Machine Learning Operations (MLOps)

Statistical Data Analysis

Image processing

Time series prediction

Bayesian Inference

Research and Development

Skills

Python

Tensorflow (low level API and Keras)

Pytorch

Docker

Airflow

MLFlow

Caffe

Scikit-Learn

SQL

Git and SVN

Bash

Java (Newie :D)

Eviews

SAS/SPSS

SNPE and TIDL (Deep learning runtimes)

Excel

Tableau

Bloomberg

Languages

English (C1)
Spanish (Mother tongue)
French (Starter)

About