Diogo Dutra

Canada       
+1 437 221-1116        

I am a Lead Data Scientist and Machine Learning engineer with +14 years of international experience in Defense and Aerospace industries.

I have an entrepreneurial vein, lots of programming skills, a neck to unlock value buried in data and a passion for product development.



Experience

Machine Learning Developer

Ford Motor Company
May 2021 - present
Canada

Research and Development of future Artificial Intelligence applications for the next generation of autonomous vehicles infotainment system at Ford's R&D laboratory.

The new applications are conceived in a series of proof-of-concepts for the self-driving car including different areas such as Deep Learning, Computer Vision, user recommendation, embedded systems and connectivity with other devices.

Lead Data Scientist

Patriot One
Jul 2019 - Jan 2021
Canada

Development of an alert system capable of covertly screening moving individuals for on-body concealed weapons.

Applied Convolution Neural Network and Digital Signal Processing on radar and magnetic signals.

Mentored Data Science co-ops from University of Toronto for data collection and data analysis.

Got promoted in 3 months to lead a technical team, reporting directly to VP of Engineering.

Technology Consultant

Altran, Natura
Apr 2017 - Jul 2019
Brazil, Portugal

Embedded code for automotive cluster: BMW and Continental.

Proofs-of-Concept for Industry 4.0 for manufacturing maintenance including Additive Manufacturing (3D printer), Augmented Reality and Artificial Intelligence.

CTO

GoEpik
Nov 2016 - Jun 2017
Brazil

Co-founder and CTO of the most attractive Brazilian startup in 2017 during my part-time by offering Augmented Reality app for maintenance.

Systems Engineer

ICN, DCNS
Apr 2013 - Mar 2017
Brazil, France

Commissioning of submarine combat system: IVVQ, STW and FAT.

Project and Commercial Manager

Mectron
Nov 2011 - Mar 2013
Brazil

$200K contract prospected and managed, ending with profit.

Aerospace Engineer

Embraer, Mectron, Denel
Aug 2007 - Oct 2011
Brazil, South Africa

Guidance, Control, Navigation for missiles applying sensor fusion with Kalman Filter.

Simulation of jet fighter engine.


Education

AI Product Manager

Udacity NanoDegree
Jan 2020 - Feb 2020

Artificial Intelligence product lifecycle, how to conceive a Business Case for Machine Learning, FigureEight online tool for data annotation, Google AutoML for fast Data Science project and online deployment.

Deep Learning

Udacity NanoDegree
Dec 2019 - Jan 2020

Convolutional Neural Network, Recurrent Neural Network, Generative Adversarial Network, Transfer Learning, deploy web app online, reading articles to implement state-of-the-art neural networks.

Machine Learning

Udacity NanoDegree
Jul 2018 - Dec 2018

Supervised, Unsupervised, Reinforcement Learning.

Data Science

edX
Aug 2017 - Dec 2017

Table merging, data cleaning, visualization, integration with Office 365.

MBA

FGV
Jan 2011 - Dec 2014

Entrepreneurship, project management, marketing.

Master in Aeronautics

ITA
Jun 2006 - Feb 2010

Aerodynamics, flight dynamics and control.

Bachelor in Mechanics

IME
Jun 2006 - Feb 2010

Structural analysis, thermodynamics, fluids, simulation and weapon design.


Skills

Programming Languages & Tools
  • Python

    Python

  • Matlab

    Matlab

  • C++

    C++

  •   
  • PyTorch

    PyTorch

  • Keras

    Keras


  • Google Cloud Platform

    Google
    Cloud Platform

  • Amazon Web Services

    Amazon
    Web Services

  • Microsot Power BI

    Microsoft
    Power BI

  •   
  • Git

    Git
     

  • JIRA

    JIRA
     


Idioms
  • English, French and Portuguese

Projects

Video Special Effects

Video Processing

I developed the core of the video processing technology for one of the Big Five Giant American tech companies. Now you can use your camera to create custom special effects such as background removal and augmented reality animations. This technique is transferable to online streaming data such as filters.

No source: original protected trade secret work.

More samples of artwork: Platinum Awards, Event ID



Country of the temple

Convolutional Neural Network

I created this classifier to tell from which country a temple, church, mosque, monastery or stupa is from. This technique is transferable to any other classification task not only for images but also audio, text and sensors as well.

Source-code

Demo



COVID-19 triage by hemogram

Volunteer Work

I created this classifier to help the hospitals to triage the suspects of COVID-19 from a simple blood test! It is faster and spares many SARS-CoV-2 test kits. It can save lives and costs at the same time during the peak of the coronavirus pandemic.

Dataset: Kaggle

Source-code

Demo



Diagnose Pneumonia

Google AutoML Cloud Services

What if we could fasten the diagnosis process? My classifier was trained with a dataset of X-ray images from children chests. Now, it labels any X-rays image as either Normal, Bacterial pneumonia or Viral pneumonia. The result is an automated diagnosis that can accelerate the clinical care for children with world-class precision.

Dataset: Kaggle



Fake Face Generator

Generative Adversarial Networks

I trained a Generative Adversarial Networks (GAN) using a dataset of celebrities photos to learn how to create a fake human face. This same technique can be used to generate any realistic content like images, texts, simulations or signals.

Paper: "Deep Cyclic Generative Adversarial Network", Facebook AI Research

Source-code here



Concealed Weapon Detector

Convolutional Neural Networks

I applied Computer Vision and Digital Signal Processing techniques on radar for classification. My model finds meaningful contrasts between weapons and decoys on a person in movement. CNN can be used to analyse not only images but also any other signals such as video, audio, radar and sonar.

No source: original protected trade secret work.



AI writter

Natural Language Processing

What if an Artificial Intelligence could kickstart an original text so you do not have to start a new one from scratch? I created this website that the user enters a few keywords and it gives back some related suggestions of text with titles. I used Heroku, RESTful and asynchronous tasks to deploy this model online.

No source: personal project deployed for commercial purpose.

Create your text here



Movie Recommender

Unsupervised Learning

I created this Collaborative Filter as a recommendation engine for a cloud-service for movie streaming. Any online business must leverage its database to find out and deploy a better user experience to drive consumption. Unsupervised Learning is ideal to find customer segments, which can be used to predict common preferences.

Paper: "Collaborative Filtering", Netflix Grand Prize

Source-code here



Dog Breed Classifier

Transfer Learning

This classifier can tell which is the breed of the dog in a picture. I used OpenCV combined with a customized ResNet50 by applying Transfer Learning in order to save time, since creating another one from scratch would take much longer to train my model. Spin-offs of this technique allow us to tell the content of any media.

Sources: "Transfer Learning", Stanford - US

Source-code here



Flight Delay Predictor

Supervised Learning

The passengers on the world's busiest airway now know if their flight is going to delay with 15 minutes of margin. I created this Supervised Learning model considering a 2-years long hourly dataset of airport departures, arrivals and weather.

Source-code here



Quadcopter Autopilot

Reinforcement Learning

The Parrot quadcopter just gained an autopilot. The Reinforcement Learning knows how much power each rotor must apply in order to stabilize the drone thanks to thousands of simulated runs. Let me know if you need specific maneuvers or an autopilot for any other vehicle.

Paper: "Deep Deterministic Policy Gradients", Google Deepmind - UK

Source-code here



Bike Sharing Demand Forecast

Recurrent Neural Network

Now, the owner of a bike sharing company from Washington can make better decisions on how many bikes to buy. This forecast helps the owner to avoid losing revenue by having too few or skyrocketing the oeprational costs by having too many. I created this model by training a RNN over a 2-year hourly rental and weather data. RNNs are suitable to predict any time-series data such as demand, weather and stock options.

Dataset: "Event labeling", University of Porto - Portugal

Source-code here



Customer Segmentation

Unsupervised Learning

A retailer from Portugal wanted to lower the frequency of deliveries to reduce cost, but without considerebly disturbing its clients. It was achieved by finding its customer segments and selecting which would not be significantly impacted with 3 weekly deliveries instead of 5. Unsupervised Learning is ideal to find customer segments and understand its specific purchase behaviours.

Source-code here



Painting Style Copier

Style Transfer

The Style Transfer technique allows the Convolutional Neural Network to learn independently the content from one picture and the style from another. Later, I gradually applied the style from the painting on the content of the picture with the octopus.

Paper: "Image Transfer Style", University of Tübingen - Germany

Source code here




Key Achievements



As Data Scientist:

  • Detected concealed weapons with radar and Convolutional Neural Network

  • Stabilized a quadcopter autopilot using Reinforcement Learning

  • Predicted flight delays with Supervised Learning within margin of 15 min

  • Reduced 30% of $5M stock value of spare parts by applying data analysis for MRO warehouse



As Aerospace Engineer:

  • Increased missile flight lifespan from 1 to 10 minutes by applying sensor fusion of INS, GPS and baroaltimeter with embedded Kalman Filter

  • Missile target position estimation using Extended Kalman Filter

  • Jet bomber engine simulation to support system-level decisions across different engineering departments



As Entrepreneur, Project and Commercial Manager:

  • Co-founder and CTO of the most attractive Brazilian startup in 2017 during my part-time by offering Augmented Reality app for maintenance

  • $200K contract prospected and managed, ending with profit


Testemonials

Diogo has been a key member and an asset for our research and development team in patriot one technologies from the moment he joined. He indeed revolutionised the efficiency and outcome of our team with his excellent problem solving, mentoring, time management, data analysis and coding skills. He is expert in taking concepts from initial prototype to a potential product. We are all pleased he is part of our company. He is really a talented, hard-working, mature, wise and easy-going person.I guarantee that he brings prosperity, new knowledge and new skill sets to any company he will work for.

- Reza Mousavi

Principal Research Scientist - Patriot One






It is with great pleasure that I am writing this recommendation for Diogo. He is very detail oriented, patient, reliable and has excellent leadership skills. His ideas had helped the team in formulating a stronger product and he is incredibly generous with his time and expertise, which includes everything from 10+ years of technical experience to an expert-level knowledge of in Design, RADAR and Machine learning models. He takes time in listening to all the opinions and try to formulate a solution which best suits the problem and benefit the company. I wish we get to work together again and wish him all the success.

- Harika Gaggara

Data Scientist co-op - Patriot One






I had the great pleasure to worked on the Natura Industry 4.0 Project, an exceptional professional, very focused on innovation, owner of incredible ideas, I hope to have the pleasure to work again on another project with Diogo!

- Alexandre Vasquez

Big Data, IoT & Advanced Analytics - Microsoft






Mr. Dutra has a strong leadership presence, being able to coordinate a team but also capable to work solo. Dutra has also a great sense of duty regarding his goals and always is concerned with the quality of the work under his responsibility.

- Raphael Calvo

Chief Engineer Office - Embraer






Diogo is a brilliant guy. With a very strong math and physics base, he is capable of doing very high quality engineering work - doesn't matter the field of knowledge. He is an easygoing person to work with and very helpful as well.

- Ricardo Mello

Flight Dynamics - MITAC






Diogo is a high skilled professional, capable of solving complex engineering problems. Very creative person, he is always updated on how to apply new technologies.

- Ricardo Esteves

Technical Manager - Avibrás






I had a pleasure to work with Diogo for 2 years. He is a high skilled professional for complex problems solutions, shaping solutions out of complexity, besides being very commited to his duties and a good partner to work.

- Lucas Supino

Commercial Manager - Thales






We work on the same team and chatted for long periods about everything, engineering, technology, hobbies and books. Diogo out stand at critical analysis and has an incredible capacity to merge different ideas to create something new. Understand the problem and solving it integrating new technologies is his best suit.

- Bruno Galbiati

Logystics Analyst - Natura






Diogo Dutra is a friend and work with me at Mectron Odebrecht Defense & Technology Company, he is a person very determined, methodical, with excellent interpersonal skills and very clear, objective in their activities, and good listener.

- Eduardo Antequera

Supply Chain - Mectron







Blog

My most important lesson after 1 year applying Machine Learning to development of radar

Combining ML with the traditional Digital Signal Processing is much more powerful than any of these isolated.

Read more

Another sad event on Artificial Intelligence ethics

YOLO creator Joseph Redmon stopped researching Computer Vision due to concerns with military and anti-privacy applications of his findings.

Read more

How simple our technical report should be?

How much we as Data Scientists / Machine Learning engineers should simplify our reports to be understood by all without hurting our reputation?

Read more

What is Deep Learning?

It is an area of Machine Learning that uses many more layers than the first generation of Neural Networks.

Read more

How Data Scientists use the same skills to solve different problems?

Elementary, my dear Watson. We reframe every problem as a prediction scenario. And reusable prediction scenarios are the Machine Learning expertise.

Read more

Want to train a Deep Learning but you got no GPU and no money?

Check these 2 options of cloud servers that are perfect to create your Machine Learning models in Python with CUDA enabled on Jupyter Notebook for free.

Read more

Does the machine really "learn by itself?"

Yesterday I was questioned "but we still need a programmer as the machines cannot learn it all by itself, right?” from my father, a veteran programmer from punched card era.

Read more

What is the mission of a Data Scientist?

As a data scientist, I couldn't be happier when I influence decision-makers to better their businesses.

Read more

"Why good leaders make you feel safe" - Simon Sinek

That is a powerful message that many of us should learn.

Read more

This is why good Data Scientists are very well remunerated

We can unlock unprecedented value buried in data, but it takes a wide variety of skills such as statistics, programming, intuition and deep understanding about the specific business area.

Read more

Bill Gates says that the only way to hire and keep AI talent is to let them share their research openly

Gates said that the US has long benefitted from openly sharing scientific research, and that it remains a huge advantage, especially within the field of artificial intelligence. In his view, there's no going back from that approach.

Read more

Agile framework for R&D

What if a Research & Development project could have the capabilities of quickly learning, pivoting and providing the information necessary for the high-level management key decisions?

Read more

What's missing for Augmented Reality to take off?

After overcoming the high cost of AR devices, Artificial Intelligence will be the next enabler for AR.

Read more