Nicolas Koumchatzky โ€” Machine Learning in Production for Self-Driving Cars

Weights & Biases ยท Intermediate ยท๐Ÿ“ ML Fundamentals ยท6y ago
๐Ÿ‘จ๐Ÿปโ€๐Ÿ’ปToday our guest is Nicolas Koumchatzky. Nicolas Koumchatzky is the Director of AI infrastructure at NVIDIA, where he's responsible for MagLev, the production-grade machine learning platform by NVIDIA. His team supports diverse ML use cases: autonomous vehicles, medical imaging, super resolution, predictive analytics, cyber security, robotics. He started as a Quant in Paris, then joined Madbits, a startup specialized on using deep learning for content understanding. When Madbits was acquired by Twitter in 2014, he joined as a deep learning expert and led a few projects in Cortex, include a real-time live video classification product for Periscope. In 2016, he focused on building an scalable AI platform for the company. Early 2017, he became the lead for the Cortex team. He joined NVIDIA in 2018. ๐ŸฆFollow Nicolas on twitter: https://twitter.com/nkoumchatzky ๐Ÿ› Maglev: https://blogs.nvidia.com/blog/2018/09/13/how-maglev-speeds-autonomous-vehicles-to-superhuman-levels-of-safety/ โœ๏ธScalable Active Learning for Autonomous Driving: https://medium.com/nvidia-ai/scalable-active-learning-for-autonomous-driving-a-practical-implementation-and-a-b-test-4d315ed04b5f โœ๏ธActive Learning โ€“ Finding the right self-driving training data doesnโ€™t have to take a swarm of human labelers: https://blogs.nvidia.com/blog/2020/01/16/what-is-active-learning/ Topics covered: 0:00 intro 0:42 Nicholas intro 0:52 how has deep learning shifted since 2016? 11:52 Surprisingly challenging things at the time 13:15 moving to NVIDIA 15:36 components of the infrastructure at NVIDIA, active learning 28:53 How do you approach makiing sophisticated systems on behalf of customers? 31:24 Synthetic Data 33:55 What is there left to do for the progress of autonomous vehicles? 38:02 Difference at approach between NVIDIA, Tesla, Lyft 41:11 More on Active Learning Weights and Biases makes developer tools for machine learning: record and visualize every detail of your research, collaborate easily, advance the
Watch on YouTube โ†— (saves to browser)
Sign in to unlock AI tutor explanation ยท โšก30

Playlist

Uploads from Weights & Biases ยท Weights & Biases ยท 42 of 60

โ† Previous Next โ†’
1 0. What is machine learning?
0. What is machine learning?
Weights & Biases
 2 1. Build Your First Machine Learning Model
1. Build Your First Machine Learning Model
Weights & Biases
 3 Intro to ML: Course Overview
Intro to ML: Course Overview
Weights & Biases
 4 2. Multi-Layer Perceptrons
2. Multi-Layer Perceptrons
Weights & Biases
 5 3. Convolutional Neural Networks
3. Convolutional Neural Networks
Weights & Biases
 6 Weights & Biases at OpenAI
Weights & Biases at OpenAI
Weights & Biases
 7 Why Experiment Tracking is Crucial to OpenAI
Why Experiment Tracking is Crucial to OpenAI
Weights & Biases
 8 4. Autoencoders
4. Autoencoders
Weights & Biases
 9 5. Sentiment Analysis
5. Sentiment Analysis
Weights & Biases
 10 6. Recurrent Neural Networks [RNNs]
6. Recurrent Neural Networks [RNNs]
Weights & Biases
 11 7. Text Generation using LSTMs and GRUs
7. Text Generation using LSTMs and GRUs
Weights & Biases
 12 8. Text Classification Using Convolutional Neural Networks
8. Text Classification Using Convolutional Neural Networks
Weights & Biases
 13 9. Hybrid LSTMs [Long Short-Term Memory]
9. Hybrid LSTMs [Long Short-Term Memory]
Weights & Biases
 14 Toyota Research Institute on Experiment Tracking with Weights & Biases
Toyota Research Institute on Experiment Tracking with Weights & Biases
Weights & Biases
 15 Weights and Biases - Developer Tools for Deep Learning
Weights and Biases - Developer Tools for Deep Learning
Weights & Biases
 16 Introducing Weights & Biases
Introducing Weights & Biases
Weights & Biases
 17 10. Seq2Seq Models
10. Seq2Seq Models
Weights & Biases
 18 11. Transfer Learning for Domain-Specific Image Classification with Small Datasets
11. Transfer Learning for Domain-Specific Image Classification with Small Datasets
Weights & Biases
 19 12. One-shot learning for teaching neural networks to classify objects never seen before
12. One-shot learning for teaching neural networks to classify objects never seen before
Weights & Biases
 20 13. Speech Recognition with Convolutional Neural Networks in Keras/TensorFlow
13. Speech Recognition with Convolutional Neural Networks in Keras/TensorFlow
Weights & Biases
 21 14. Data Augmentation | Keras
14. Data Augmentation | Keras
Weights & Biases
 22 15. Batch Size and Learning Rate in CNNs
15. Batch Size and Learning Rate in CNNs
Weights & Biases
 23 Applied Deep Learning Fellowship Overview and Project Selection with Josh Tobin (2019)
Applied Deep Learning Fellowship Overview and Project Selection with Josh Tobin (2019)
Weights & Biases
 24 Grading Rubric for AI Applications with Sergey Karayev (2019)
Grading Rubric for AI Applications with Sergey Karayev (2019)
Weights & Biases
 25 16. Video Frame Prediction using CNNs and LSTMs (2019)
16. Video Frame Prediction using CNNs and LSTMs (2019)
Weights & Biases
 26 Image to LaTeX - Applied Deep Learning Fellowship (2019)
Image to LaTeX - Applied Deep Learning Fellowship (2019)
Weights & Biases
 27 17. Build and Deploy an Emotion Classifier (2019)
17. Build and Deploy an Emotion Classifier (2019)
Weights & Biases
 28 Applied Deep Learning - Data Management with Josh Tobin (2019)
Applied Deep Learning - Data Management with Josh Tobin (2019)
Weights & Biases
 29 Snorkel: Programming Training Data with Paroma Varma of Stanford University (2019)
Snorkel: Programming Training Data with Paroma Varma of Stanford University (2019)
Weights & Biases
 30 Applied Deep Learning - Troubleshooting and Debugging with Josh Tobin (2019)
Applied Deep Learning - Troubleshooting and Debugging with Josh Tobin (2019)
Weights & Biases
 31 Troubleshooting and Iterating ML Models with Lee Redden (2019)
Troubleshooting and Iterating ML Models with Lee Redden (2019)
Weights & Biases
 32 Designing a Machine Learning Project with Neal Khosla (2019)
Designing a Machine Learning Project with Neal Khosla (2019)
Weights & Biases
 33 Lukas Beiwald on ML Tools and Experiment Management (2019)
Lukas Beiwald on ML Tools and Experiment Management (2019)
Weights & Biases
 34 Building Machine Learning Teams with Josh Tobin (2019)
Building Machine Learning Teams with Josh Tobin (2019)
Weights & Biases
 35 Pieter Abeel on Potential Deep Learning Research Directions (2019)
Pieter Abeel on Potential Deep Learning Research Directions (2019)
Weights & Biases
 36 Testing and Deployment of Deep Learning Models with Josh Tobin (2019)
Testing and Deployment of Deep Learning Models with Josh Tobin (2019)
Weights & Biases
 37 Five Lessons for Team-Oriented Research with Peter Welder (2019)
Five Lessons for Team-Oriented Research with Peter Welder (2019)
Weights & Biases
 38 Applied Deep Learning - Rosanne Liu on AI Research (2019)
Applied Deep Learning - Rosanne Liu on AI Research (2019)
Weights & Biases
 39 Making the Mid-career Leap from Urban Design to Deep Learning/Data Science
Making the Mid-career Leap from Urban Design to Deep Learning/Data Science
Weights & Biases
 40 Organizing ML projects โ€” W&B walkthrough (2020)
Organizing ML projects โ€” W&B walkthrough (2020)
Weights & Biases
 41 Brandon Rohrer โ€” Machine Learning in Production for Robots
Brandon Rohrer โ€” Machine Learning in Production for Robots
Weights & Biases
 โ–ถ Nicolas Koumchatzky โ€” Machine Learning in Production for Self-Driving Cars
Nicolas Koumchatzky โ€” Machine Learning in Production for Self-Driving Cars
Weights & Biases
 43 My experiments with Reinforcement Learning with Jariullah Safi
My experiments with Reinforcement Learning with Jariullah Safi
Weights & Biases
 44 Applications of Machine Learning to COVID-19 Research with Isaac Godfried
Applications of Machine Learning to COVID-19 Research with Isaac Godfried
Weights & Biases
 45 Testing Machine Learning Models with Eric Schles
Testing Machine Learning Models with Eric Schles
Weights & Biases
 46 How Linear Algebra is not like Algebra with Charles Frye
How Linear Algebra is not like Algebra with Charles Frye
Weights & Biases
 47 Predicting Protein Structures using Deep Learning with Jonathan King
Predicting Protein Structures using Deep Learning with Jonathan King
Weights & Biases
 48 Rachael Tatman โ€” Conversational AI and Linguistics
Rachael Tatman โ€” Conversational AI and Linguistics
Weights & Biases
 49 Reformer by Han Lee
Reformer by Han Lee
Weights & Biases
 50 Sequence Models with Pujaa Rajan
Sequence Models with Pujaa Rajan
Weights & Biases
 51 GitHub Actions & Machine Learning Workflows with Hamel Husain
GitHub Actions & Machine Learning Workflows with Hamel Husain
Weights & Biases
 52 Look Mom, No Indices! Vector Calculus with the Frรฉchet Derivative by Charles Frye
Look Mom, No Indices! Vector Calculus with the Frรฉchet Derivative by Charles Frye
Weights & Biases
 53 Jack Clark โ€” Building Trustworthy AI Systems
Jack Clark โ€” Building Trustworthy AI Systems
Weights & Biases
 54 Surprising Utility of Surprise: Why ML Uses Negative Log Probabilities - Charles Frye
Surprising Utility of Surprise: Why ML Uses Negative Log Probabilities - Charles Frye
Weights & Biases
 55 Track your machine learning experiments locally, with W&B Local - Chris Van Pelt
Track your machine learning experiments locally, with W&B Local - Chris Van Pelt
Weights & Biases
 56 Antipatterns in open source research code with Jariullah Safi
Antipatterns in open source research code with Jariullah Safi
Weights & Biases
 57 Attention for time series forecasting & COVID predictions - Isaac Godfried
Attention for time series forecasting & COVID predictions - Isaac Godfried
Weights & Biases
 58 Made with ML - Goku Mohandas
Made with ML - Goku Mohandas
Weights & Biases
 59 Angela & Danielle โ€” Designing ML Models for Millions of Consumer Robots
Angela & Danielle โ€” Designing ML Models for Millions of Consumer Robots
Weights & Biases
 60 Deep Learning Salon by Weights & Biases
Deep Learning Salon by Weights & Biases
Weights & Biases

Related AI Lessons

โšก
The Idempotency Nightmare in AI Pipelines: Data Loss and Recovery
Learn how idempotency issues can cause data loss in AI pipelines and how to resolve them
Dev.to AI
 โšก
My Experience with Network Anomaly Detection Using 5 Different ML Approaches
Learn from a developer's experience with network anomaly detection using 5 different ML approaches to improve your skills in machine learning and network security
Medium ยท Machine Learning
 โšก
My Experience with Network Anomaly Detection Using 5 Different ML Approaches
Learn from a developer's experience with 5 different ML approaches for network anomaly detection and improve your own detection skills
Medium ยท Cybersecurity
 โšก
Sujar Henry on Why Access Still Isnโ€™t Enough in Tech
ML expert Sujar Henry emphasizes that access to tech isn't enough, beginners need a clear path to follow
Medium ยท Machine Learning

Chapters (11)

intro
0:42 Nicholas intro
0:52 how has deep learning shifted since 2016?
11:52 Surprisingly challenging things at the time
13:15 moving to NVIDIA
15:36 components of the infrastructure at NVIDIA, active learning
28:53 How do you approach makiing sophisticated systems on behalf of customers?
31:24 Synthetic Data
33:55 What is there left to do for the progress of autonomous vehicles?
38:02 Difference at approach between NVIDIA, Tesla, Lyft
41:11 More on Active Learning
Up next
Generative Artificial Intelligence Full Course 2026 | Gen AI Tutorial For Beginners | Simplilearn
Simplilearn
Watch โ†’