Amazon re:MARS 2022
2022-06-21..2022-06-24
Las Vegas, United States
| title | author | length |
|---|---|---|
| Next-generation tools for synthetic data and AI training | Nathan Kundtz | 33m |
Platform Rendered.ai provides a full synthetic-data engineering stack: asset ingestion, Blender/Omniverse/DIRSIG/Quadridox integrations, cloud GPU… | ||
| Open space: A revolution in robots for space exploration | Michael Jeronimo | 51m |
Robotics on missions already run real systems: Lunokhod 1 lasts 321 days, Lunokhod 2 drives 39 km, Perseverance ships ~2 million lines of code, and Ingenuity… | ||
| Building virtual worlds for simulation with Unreal Engine & AWS Ambit | Aaron Sternberg, Kris Schultz | 40m |
AWS shows a concrete workflow that turns OpenStreetMap data and Unreal assets into ready-to-run simulator scenarios using the AWS Ambit Scenario Designer… | ||
| Viasat and AWS: Resilient tactical edge capability | Phil Mar | 36m |
Viasat and AWS show a deployable tactical edge that runs real time image recognition from a truck and aircraft using AWS Snowball Edge plus cloud training on… | ||
| Monitoring global events with AI-driven satellite swarms | Scott Herman | 49m |
Claims an AI brain that uses reinforcement learning to auto-orchestrate swarms of SAR, multispectral and RF satellites, running on AWS and modeling… | ||
| How deep learning can help interpret American Sign Language | Ashwini Rudra, Sarita Joshi, Arnab Ghosh | 35m |
Prototype converts static ASL finger spelling to text using transfer learning on SageMaker Studio and MXNet, training on about 72K images across 29 classes… | ||
| How Miso Robotics built an AI-activated robot fry cook | Chris Kruger | 28m |
Flippy is a rail-mounted robot fry cook that has cooked over 50,000 baskets at a White Castle install, built with ROS, eight cameras and on-site edge compute… | ||
| ML methods to accelerate the drug discovery cycle | Erika Pelaez Coyotl (she/her), Zichen Wang (he/him), Peter Clark (he/him) | 31m |
Built a Janssen knowledge graph on AWS that links molecules, assays, lab notebooks and regulatory docs using Amazon Neptune so teams query unified records… | ||
| Interplanetary compute: The Moon, Mars, and beyond | Luca Rossettini, Fredrik Bruhn, Buffy Wajvoda | 56m |
Claim: put cloud compute into orbit to cut launch-to-revenue and downlink costs. Hardware: X86 CPUs, AMD GPUs, Intel Myriad VPUs, FPGAs running… | ||
| Natural and conversational shopping experiences | Ujjwal Ratan, Vidya Sagar Ravipati | 35m |
Ambient Explorer stitches web crawling, transformer-based information extraction, neural document summarization and a multimodal neural retrieval engine to… | ||
| Teaching a robot new tricks | Marc Theermann, Brad Bonn, Jon Slominski | 01h 00m |
Spot operates as a deployable inspection platform: about 1,000 units sold, ~$100,000 each, carries ~30 lb payloads and runs ~90 minutes battery with docking… | ||
| Machine learning for computer vision in pharmaceutical manufacturing | Thomas Bach Nielsen, Jonas Vejlgård Kristensen, Carlos Ribera Codina | 30m |
Claim of an industry-first automated quality assurance system that counts about 370 cartridges per box in real time using a robotic arm, top-down camera rig… | ||
| Improve explainability of ML models to meet regulatory requirements | Suraj Muraleedharan | 44m |
Explains how Amazon SageMaker Clarify runs explainability and bias checks across the ML lifecycle, producing S3/PDF reports, CloudWatch metrics and alerts, and… | ||
| Innovating power plant operations through digital twins | Adam Rasheed, Clarissa Raithel | 59m |
Explains a working digital twin pipeline that uses AWS IoT TwinMaker for 3D context, Amazon Lookout for Equipment for unsupervised anomaly detection, and the… | ||
| Creating high-quality training sets for machine learning models | Morgan Dutton, Hermann Burgmeier | 22m |
SageMaker Ground Truth adds synthetic-data generation to produce labeled computer vision training sets using a library of digital assets and scene controllers… | ||
| Quantum machine learning: Prospects and challenges | Sohum Thakkar | 44m |
Gives concrete algorithms and results: circuits to load vectors and compute dot products by loading then reversing the loader, a MNIST nearest-centroid… | ||
| Accelerating AI & ML at scale with Teradata Vantage & SageMaker | Trish Lugtu, M.S., Kris Skrinak, Scott DuPertuis | 56m |
Teradata Vantage integrates with Amazon SageMaker to keep feature engineering and large-scale data prep inside a massively parallel database while training and… | ||
| Accelerate autonomous vehicle development | Cher Simon, Chris Rampley | 40m |
Capgemini’s DASV on AWS claims to cut time to data by moving from a five‑day upload delay to near real‑time streaming using AWS Wavelength, Snowball, Outposts… | ||
| Space 2050 | Nelson Pedreiro | 59m |
Lockheed Martin outlines a concrete Space 2050 roadmap: automate design to “one‑click” and mass‑customize spacecraft, push rapid prototyping and frequent… | ||
| Functional safety product development for autonomous mobile robots | Nia Jetter, Justin Croyle | 51m |
A functional safety process maps HARA-derived safety functions to design, test and operations and is already applied to three Amazon AMR products; the team… | ||
| Redefine how enterprises work w/ Blue Prism intelligent automation | Jon Day, Gregg Brett, Madhu Raman | 01h 00m |
Demonstrates an AWS + Blue Prism stack that combines RPA, ML (Rekognition, Comprehend, Textract), process mining and task mining to discover and drive… | ||
| Extract information accurately from documents using computer vision | Navneeth Nair, Martin Schade | 42m |
Amazon Textract extracts printed text, handwriting, tables and key‑value pairs and adds two new APIs (Analyze Expense, Analyze ID) plus a natural‑language… | ||
| The mathematics of machine learning | Ian Deters | 56m |
Claim: the target conditional expectation is (for practical purposes) a continuous function on the unit hypercube, so any family that densely spans continuous… | ||
| Net zero carbon goal and Amazon’s fulfillment network | Loan T. Le | 26m |
Measured every motor drive in a robotics-sortable facility and finds conveyors move about 500 tons/day while total motor power equals a 100‑car freight train… | ||
| AI for making decisions in clinical medicine | Pranav Rajpurkar | 47m |
Describes building self‑supervised models that pretrain on unlabeled medical images and audio then fine‑tune on small labeled sets, demonstrates zero-shot… | ||
| Sharing solar energy through peer-to-peer microgrids in Bangladesh | Munir Hasan, Hannes Kirchhoff | 12m |
SOLShare builds peer-to-peer village microgrids that let households bank and sell excess solar via mobile-money wallets, using a field device based on ESP32… | ||
| Disrupting new space with real-time dynamic monitoring | Patrick O'Neil | 49m |
BlackSky claims operational real‑time geospatial intelligence: a Spectra AI SaaS ingesting ~1M observations/day from 100k+ sources and distilling ~18,000… | ||
| Using machine learning to improve player safety in the NFL | Rob Mulla, Jarvis Lee, Betty Zhang | 33m |
A machine pipeline combines helmet object detection, multi-view tracking, and NGS sensor alignment to assign impacts to players, then classifies 16-frame… | ||
| Computer vision for automated quality inspection | Prathyusha Cheruku, Gaurav Kaila | 27m |
Claims that Amazon Lookout for Vision builds usable custom defect models as few as 30 images, runs inference on-device like an Nvidia Jetson edge inference… | ||
| Extracting real-time data at the edge with SageMaker & NVIDIA GPUs | Ryan Vanderwerf, Eddie Weill, Yuxin Yang | 45m |
Pipeline shows a people‑counting system that compiles a YOLOv3 pupil detector with SageMaker Neo, optimizes with TensorRT, and deploys a Neo/TensorRT artifact… | ||
| Predicting failure & automating defect detection on utility assets | Jade-Bianca Thomas, Tyler McKerihan, Cooper Wakefield | 39m |
PIPE AI deploys a cloud pipeline on AWS (ECS, EFS, Lambda, Direct Connect/PrivateLink) that automates CCTV review with computer vision and a Review and Predict… | ||
| Advance diversity & inclusion in voice AI with speech disentanglement | Ewa Kolczyk, Kayoko Yanagisawa | 48m |
Amazon describes concrete ML work: a neural TTS pipeline with encoder‑attention‑decoder and vocoder that adds speaker ID, language ID, accent ID to produce a… | ||
| Human-like reasoning for an AI | Craig Saunders, Devesh Pandey | 30m |
Amazon is building Alexa features that combine extracted events, a knowledge graph and stateful smart‑home signals to reason across context. Systems use web… | ||
| The democratization of tomorrow’s technology | Sree Ganesan, Jonathan Luse, Erik Nieves | 50m |
Intel pitches Gaudi DL1 and Gaudi2 as purpose-built training accelerators claiming up to 40–77% better price/performance versus GPUs and nearly 2x throughput… | ||
| Automation: Amazon’s Supply Chain Optimization Technologies (SCOT) | Deepak Bhatia, Salal Humair | 01h 00m |
A large-scale automated system called SCOT runs buy/place/replenish/fulfill decisions for tens of millions of SKUs worldwide, using probabilistic forecasts… | ||
| Why Airflow is a secret ingredient in MLOps | Steven Hillion | 56m |
Claims Airflow is the orchestration backbone for MLOps, with Astronomer contributing ~50% of commits and Airflow seeing ~9 million downloads/month. Real… | ||
| How Amazon builds AI-enabled perception for robots | Bhavana Chandrashekhar | 29m |
Robin is a package manipulation robot that segments and classifies cluttered conveyor scenes using an instance segmentation model and a defect classification… | ||
| Leveraging AI to optimize space resources and drive autonomy | Luca Rossettini, Meghan Weber | 40m |
Company runs a cargo vehicle called ION cargo that deploys multiple satellites to target orbits in two to four weeks, cutting time-to-revenue by up to 85% and… | ||
| Modeling our future: Advancing climate research with AI | Kelly Wanser, Kalai Ramea, Elizabeth A. Barnes | 49m |
Neural nets predict maps’ year from climate-model snapshots and Explainable AI backpropagates to show which regions drive that signal (Southern Ocean, North… | ||
| How voice agents engage customers in multimodal interactions | Ruhi Sarikaya | 29m |
Blended results merges first-, second-, and third‑party skills on multimodal devices to present multiple answers instead of a single reply, e.g., “order pizza”… | ||
| Reducing the environmental impact of agriculture with robotics | Gareth Edwards | 33m |
Robotti is a 3.4‑ton autonomous implement handler (two 150 Kubota engines) deployed as 30+ robots across four continents, using RTK GPS and sub-inch precision… | ||
| Delivering content directly, without the need for an antenna | Alistair McLean, Jason Rizzo | 13m |
SES routes satellite IP feeds straight into AWS so production and processing run in the cloud instead of at a ground station, replacing antenna downlinks with… | ||
| Managed federated learning on AWS: A case study for healthcare | Ujjwal Ratan, Vidya Sagar Ravipati | 45m |
Reproduces federated training on eICU (200k stays, ~200 hospitals) to predict in‑hospital mortality, compares centralized vs federated runs, and evaluates open… | ||
| Self-learning Alexa: ML model updates with no human in the loop | Aaron Eakin, Angela Sun | 52m |
A real-time continual learning system labels live Alexa audio with semi‑, weak‑ and self‑supervised methods (teacher→student distillation, noisy‑student… | ||
| Taking the planet’s pulse: AI to empower adaptation to global change | Joseph O. Sexton, PhD (he/him), Min Feng, PhD (he/him), Ana Pinheiro Privette, PhD | 48m |
Amazon hosts a cloud program that places petabytes of Landsat, Sentinel and other satellite data in AWS and terraPulse uses that stack plus GEDI lidar and… | ||
| Optimizing AI/ML workloads for sustainability | Amogh Gaikwad | 32m |
Cloud guidance focuses on measurable savings: AWS claims AWS is 3.6x more energy efficient and 88% lower carbon versus median enterprise datacenters… | ||
| Business transformation through data and ML | Stepan Pushkarev, Russell Lamb | 53m |
PepsiCo ecommerce uses AWS and Provectus to build a cloud data platform (S3 Parquet, EKS/Kubernetes, Terraform, Airflow, DBT, Lambdas, SNS, Apache Superset)… | ||
| Inside the fastest-growing trends in MLOps: VC & startup perspective | Ari Kalfayan | 59m |
Panelists map concrete MLOps moves: OctoML compiles models into portable, hardware‑optimized artifacts and plugs into CI/CD; Weights & Biases supplies… | ||
| How small robotics teams innovate and scale quickly | Thomas Ryden | 39m |
MassRobotics reports specifics: robotics market jumps from roughly $30–50B today to about $200B by 2025, cobots are only 3–5% of industrial arms though UR has… | ||
| Advancing underwater exploration with robotics | Marcos Barrera | 42m |
Seafloor Systems deploys 3,000+ autonomous robots and claims quantifiable gains: HyDrone keeps transects to less than 10 ft cross-track error versus ~100 ft… | ||
| Robot trajectory planning with hybrid quantum algorithms | Kyle Brubaker | 43m |
AWS and BMW build an end-to-end robot trajectory pipeline for sealant robots that compares a QUBO→QBSolv→D‑Wave annealing path against a native random‑key… | ||
| Empowering intelligent edge IoT devices | Finbarr Moynihan | 23m |
MediaTek unveils a Genio platform for intelligent edge devices centered on the Genio 1200 SoC: 6 nm Octa‑core (4x A78 + 4x A55), Mali‑G57 MC5 GPU, dual‑core… | ||
| Improving disaster response with machine learning | Antje Barth | 58m |
A community hackathon yields 1,400 participants, 42 submissions and five winners that actually ship reproducible code and data: examples include a… | ||
| A/B testing with real-world data to evaluate ML model updates | Riaz Liyakath, Anu Jayanthi | 42m |
AWS Activate builds a personalization recommender that compares a legacy OpenSearch rules engine (control) against two ML-driven variants, using CloudWatch… | ||
| Exploratory data analysis and automated feature engineering | Isha Dua, Shreyas Subramanian | 44m |
Overview of Amazon SageMaker tools that turns manual EDA into clickable workflows: SageMaker Data Wrangler connects to S3, Athena, Snowflake or Databricks… | ||
| Robots and assistive technologies | Cam Buscaron, Sebastian Arias | 36m |
Robotics for assisted living shows concrete outcomes: a deployed Aby companion robot operating at Presbyterian Village since December 2021 with three units, 2… | ||
| Retrieving objects with voice: A new generation of assistive robots | Mike Dooley, Nikolai Romanov | 58m |
A compact Alexa-enabled service robot actually fetches and carries items (laundry baskets up to ~20–25 lbs), navigates homes with 3D stereo vision and… | ||
| Delivering the promise of conversational AI with Connect | Yoni Barkan, Daniel Spirn | 55m |
TrueServe shows a contact-center system that combines deterministic flows with probabilistic NLUs, demoing a 401(k) assistant that answers fact queries with… | ||
| The future of software development, powered by machine learning | Atul Deo, Parminder Bhatia | 22m |
AWS describes a suite: CodeGuru and DevOps Guru for reviews and profiling, and a newly announced CodeWhisperer that generates multi-line code recommendations… | ||
| nothing matches | ||