Servicios de IA en AWS

Transcript

1. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Architecting for Machine Learning J u l i o M . F a e r m a n @ j u l i o a w s
2. 20 years of Machine Learning At Amazon… Search & Discovery Fulfilment & Logistics Existing Products New Initiatives
4. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
5. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
6. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
7. Amazon Prime Air
8. Amazon Go
10. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Put machine learning in the hands of every developer and data scientist ML @ AWS: Our mission
13. “With Amazon Polly our users benefit from the most lifelike Text-to-Speech voices available on the market.” Severin Hacker CTO, Duolingo
14. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. CUSTOMERS RUNNING ML ON AWS
15. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
16. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. ML Infrastructure
17. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. P3 Instances • Up to eight NVIDIA Tesla V100 GPUs • 1 PetaFLOP of computational performance – 14x better than P2 • 300 GB/s GPU-to-GPU communication (NVLink) – 9X better than P2 • 16 GB GPU memory with 900 GB/sec peak GPU memory bandwidth
18. F1 Instances
19. MQTT AWS IoT 1-click Endpoints Gateway/PLC Cloud Enterprise Applications Device shadow Rules Engine AWS IoT Core Certificate Authority Device Shadow AWS Greengrass Lambda Functions Message Router Local Comms Long-range Comms Amazon FreeRTOS Certificate Authority Local Resourc es OPC-UA Adapter IoT SDK AWS IoT Device Management AWS IoT Analytics AWS Amazon Kinesis Enterprise Users Corp Apps Amazon QuickSight Amazon EMR Amazon Redshift Amazon S3 Machine Learning OPC-UA MQTT MQTT IoT Users Edge Users IoT Partners Cer t AWS Lambda WiFi All AWS Over-the-air (OTA) Updates Analytics Data Store Data Pipelines Templated Reports Local Resources IoT with AWS Batch Fleet Provisioning Real-time Fleet Index & Search AWS IoT Device Defender Ad-hoc & In- depth Analysis Risk Mitigation Edge ARM, Broadcom, Digi, Expressif, Intel, MediaTek, Microchip, NXP, ST, TI, Qualcomm, … Gateway Adlink Technology, Advantech, MachineShop, Technicolor, … Detection Profiles Alerts Scheduled or Ad-hoc Audit MQTT Platform Ayala, Bright Wolf, BSquare, C3IoT, Mnubo, PTC ThingLogix, Splunk, … Connectivity Amdocs, Asavie, AT&T, Eseye, Soracom, TATA Communications Verizon, … Consulting / ISVs Accenture, Aricent, ClearScale, CTP, Luxoft, Mobiquity, solstice, Storm Reply, Sturdy Networks, TCS, TrekIO, … Things OTA OTA Amazon FreeRTOS Message Broker Integrated Client
20. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. AWS Greengrass ML Inference(Pre v ie w ) Run machine learning models on the edge Local ML inferencing Built in package for Apache MXNet Access models trained in the cloud Access GPUs and FPGAs
21. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. ML Libraries and Services
24. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Training Deep Learning Models Scalability Flexibility vs
25. Simple, Easy-to-Understand Code Flexible, Imperative Structure Dynamic Graphs High Performace github.com/gluon-api
29. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. ML Services
30. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Data wrangling • Setup and manage notebook environments • Get data to notebooks securely Experimentation • Setup and manage clusters • Scale/distribute ML algorithms Deployment • Setup and manage inference clusters • Manage and auto scale inference APIs • Testing, versioning, and monitoring Fetch data Clean & format data Prepare & transform data Train model Evaluate model Integrate with prod Monitor/ debug/refresh
31. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. End-to-End Machine Learning Platform Zero setup Flexible Model Training Pay by the second Amazon SageMaker Build, train, and deploy machine learning models at scale $
32. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Highly-optimized machine learning algorithms Amazon SageMaker Build Pre-built notebook instances
33. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Highly-optimized machine learning algorithms Amazon SageMaker One-click training for ML, DL, and custom algorithms Build Pre-built notebook instances Easier training with hyperparameter optimization Train
34. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. One-click training for ML, DL, and custom algorithms Easier training with hyperparameter optimization Highly-optimized machine learning algorithms Amazon SageMaker Deployment without engineering effort Fully-managed hosting at scale Build Pre-built notebook instances Deploy Train
35. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Streaming datasets, for cheaper training Train faster, in a single pass Greater reliability on extremely large datasets Choice of several ML algorithms Better algorithms
36. Recommendations Amazon Machine Learning
37. Regression Year Area Rooms Garages Sold by 2016 130 m2 2 2 R$ 150.000,00 2016 100 m2 1 0 R$ 100.000,00 2017 110 m2 2 1 R$ 200.000,00 ... … … … … 2017 140 m2 3 2 ?
38. Categories Classification
39. Scikit-Learn Kaggle #1 Benchmark AML Multi-class Logarithmic Loss 1.50241 0.38899 0.63462
40. Binary classification
49. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Images and Video
50. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Launch Customers
51. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon SageMaker: Launch Customers
52. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon SageMaker: Launch Customers
53. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon SageMaker: Launch Customers
54. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Object and Scene Detection Facial Analysis Face Comparison Facial Recognition Celebrity Recognition Image Moderation Amazon Rekognition Image: Deep learning-based image analysis service
55. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Results: | IT’S - 97% | | MONDAY – 99% | |but – 97% |keep – 96% | | Smiling – 99% | DetectText Rekognition: Text in Image
56. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Real-time face recognition against tens of millions of faces <0.5 second response time Up to 10M faces Enable Immediate response Rekognition: Real-Time Face Search
57. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Rekognition Image Customers
58. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Videos != Frames Temporal information lost Motion context lost
59. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Rekognition Video Video Analysis
60. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Rekognition Video Stored Video Amazon S3 Video Live Stream Amazon Kinesis Video Stream Media Search Index Unsafe Video Detection Investigative Analysis Public Safety Immediate Response Home Monitoring
61. Media and Entertainment Public Safety Smart Home - Search & Filter - Immediate Response - Investigative Analysis - Monitoring Sample Use Cases
62. Amazon confidential Rekognition Video: Demo
63. AWS DeepLens Pre-order for April 14, 2018
64. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. AWS DeepLens: How Does It Work? Deep learning models run in real time on the device Deploy pre-trained educational models Deploy custom models from Amazon SageMaker
65. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. AWS DeepLens: How Does It Work? Deep learning models run in real time on the device Stream live video back to AWS Integrate with Lambda, SNS, SES, etc. Video analytics with Amazon Rekognition Video Sophisticated inference with Amazon IM hosted models
66. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Rekognition Video: Launch Customers
67. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Audio and Language
68. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Polly Use Cases Content Creation Mobile & Desktop Applications Internet of Things (IoT) Education & E-Learning Telephony Game Development Key Features 52 Voices across 25 languages Lip-Syncing & Text Highlighting Fine-grained Voice Control Custom Vocabularies Available in 14 AWS regions
70. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Polly Customers
71. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Transcribe: Automatic Speech Recognition (Preview) Time Stamps Support for both regular & telephony audio Punctuation & formatting § S3 Integration Recognize Multiple Speakers Custom Vocabulary
72. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Call Centers Subtitles for VOD Transcribe meetings Broadcast closed captions Amazon Transcribe: Use Cases
73. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Transcribe: Launch Customers
74. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Frameworks Infrastructure AWS Deep Learning AMI GPU (P3 Instances) Mobile CPU IoT (Greengrass) Language: Lex Speech: Amazon Polly Transcribe Vision: Rekognition Image Rekognition Video Apache MXNet PyTorch Cognitive Toolkit Keras Caffe2 & Caffe TensorFlow Gluon Application Services Platform Services AWS ML Stack Amazon Machine Learning Mechanical Turk Spark & EMR Amazon SageMaker AWS DeepLens
76. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Lex
77. Amazon Lex Utterances Spoken or typed phrases that invoke your intent BookHotel Intents An Intent performs an action in response to natural language user input Slots Slots are input data required to fulfill the intent Fulfillment Fulfillment mechanism for your intent
78. “Book a Hotel” Book Hotel NYC “Book a Hotel in NYC” Automatic Speech Recognition Hotel Booking New York City Natural Language Understanding Intent/Slot Model Utterances Hotel Booking City New York City Check In Nov 30th Check Out Dec 2nd “Your hotel is booked for Nov 30th” Polly Confirmation: “Your hotel is booked for Nov 30th” “Can I go ahead with the booking? a in
80. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Lex – Use Cases Informational Bots Chatbots for everyday consumer requests Application Bots Build powerful interfaces to mobile applications • News updates • Weather information • Game scores …. • Book tickets • Order food • Manage bank accounts …. Enterprise Productivity Bots Streamline enterprise work activities and improve efficiencies • Check sales numbers • Marketing performance • Inventory status …. Internet of Things (IoT) Bots Enable conversational interfaces for device interactions • Wearables • Appliances • Auto …. Contact Center Bots Chatbots for customer service IVR • Account inquiries • Bill payment • Service update ….
81. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Lex Customers
82. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. REAL-TIME TRANSLATION POWERED BY DEEP LEARNING 12 LANGUAGE PAIRS (more to come) LANGUAGE DETECTION Amazon Translate
83. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. A “POCKET” KNIFE ON AMAZON.DE
84. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. A HAPPY CUSTOMER WROTE: Really a very good Pocket knife. Had yesterday after breakfast something poppy hang between the teeth. Because I once again found no toothpicks, I ordered mine without further ADO this pocket knife. The integrated toothpick (Nr. 3) is very stable and very quickly I could clean my teeth. Super! To the rest I can tell nothing unfortunately because I him not be emergency.
85. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. THE NEURAL VERSION “Really a very good pocket knife. Had a little poppy between the teeth yesterday after breakfast. Since I didn’t find a toothpick again, I ordered this pocket knife shortly. The integrated toothpick (No.3) is very stable and I could clean my teeth very quickly. Great! I can’t say anything about the rest because I don’t need it.”
86. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Translate: Launch Customers
87. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Comprehend Natural Language Processing Sentiment Entities Languages Key phrases Topic modeling
88. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. A m a z o n . c o m , I n c . i s l o c a t e d i n S e a t t l e , W A a n d w a s f o u n d e d J u l y 5 t h , 1 9 9 4 b y J e f f B e z o s . O u r c u s t o m e r s l o v e b u y i n g e v e r y t h i n g f r o m b o o k s t o b l e n d e r s a t g r e a t p r i c e s N a m e d E n t i t i e s • A m a z o n . c o m : O r g a n i z a t i o n • S e a t t l e , W A : L o c a t i o n • J u l y 5 t h , 1 9 9 4 : D a t e • J e f f B e z o s : P e r s o n K e y p h r a s e s • O u r c u s t o m e r s • b o o k s • b l e n d e r s • g r e a t p r i c e s S e n t i m e n t • P o s i t i v e L a n g u a g e • E n g l i s h
89. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. F u l l y m a n a g e d T o p i c m o d e l i n g s e r v i c e E x t r a c t u p t o 1 0 0 t o p i c s f r o m a c o r p u s o f d o c u m e n t s A u t o m a t i c a l l y o r g a n i z e s d o c u m e n t s i n t o t h e t o p i c s Amazon Comprehend: Topic Modeling
90. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Comprehend: Use Cases Voice of Customer Analytics Semantic Search Knowledge Management/Discovery Analyzing what customer are saying about your brand, products, and services Making search smarter by searching on keyphrase, sentiment, and topic Organizing documents, categorizing by topic and personalizing experiences
91. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Comprehend: Launch Customers
92. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Multilingual Social Analytics Twitter Stream API Kinesis Lambda S3 Athena Translate Comprehend Transcribe
95. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Artificial Artificial Intelligence
97. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Mechanical Turk • The first and foremost challenge in build AI systems is to build ground truth data • Need human intelligence to annotate speech, vision, or language data sets
98. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Mechanical Turk ML Customers Facebook AI Research (FAIR) DigitalGlobe | Radiant
99. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Frameworks & Infrastructure AWS Deep Learning AMI GPU (P3 Instances) Mobile CPU IoT (Greengrass) Vision: Rekognition Image Rekognition Video Speech: Amazon Polly Transcribe Language: Lex Translate Comprehend Apache MXNet PyTorch Cognitive Toolkit Keras Caffe2 & Caffe TensorFlow Gluon AWS ML Stack Application Services Platform Services Amazon Machine Learning Mechanical Turk Spark & EMR Amazon SageMaker AWS DeepLens
100. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Support and Incentive
101. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. • Accelerate the development of cutting-edge ML research (across speech, language, decision making, problem solving, translation, and computer vision) • Fund with cash and AWS credits to lead to novel artificial intelligence research AWS ML Research Grants …and many more
102. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon ML Lab Lots of companies doing Machine Learning Unable to unlock business potential Brainstorming Modeling Teaching Lack ML expertise Leverage Amazon experts with decades of ML experience with technologies like Amazon Echo, Amazon Alexa, Prime Air, and Amazon Go Amazon ML Lab provides the missing ML expertise
103. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon ML Lab Customers
104. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Annotation, Wrangling, Procurement Media NLP Optimization Other SaaS/API Data Annotations Generation ‘Wrangling’ Data Management Predictive Model Training Model Evaluation Model Deployment Model Management Offline and Online Prediction Computer Vision Natural Language Processing Recommendation Engines Conversational Interfaces Event Prediction Anomaly Detection Amazon ML Competency for Technology Partners Data Science & Machine Learning Platforms
105. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved. New Solutions in AWS Marketplace D a t a S o l u t i o n s M L & D a t a S c i e n c e I n t e l l i g e n t S o l u t i o n s aws.amazon.com/mp/ai
106. https://aws.amazon.com/twitch/
107. C O N S T R U Y A M O S Julio M. Faerman @julioaws