tag:speakerdeck.com,2005:/amitkaps tag:speakerdeck.com,2005:Talk/648901 2020-06-30T12:55:37-04:00 2020-07-11T12:20:17-04:00 A perspective on how to make deep learning models more accessible and usable for a wider spectrum of people. Using tensorflow.js 1. Educate: Explorable Explanations 2. Create: Rapid Prototyping 3. Visualise: Model Visualisation 4. Intervene: Decision Making 5. Imagine: Generative Spaces Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/480450 2018-12-06T06:33:38-05:00 2020-06-30T18:12:17-04:00 Visual data stories can enable us to move from analysis to synthesis, from numbers to visuals, and from an argument to a story. In this talk I talk about the purpose of data visualisation, the principles of storytelling and how we can combine the two together to craft visual stories with data. Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/445715 2018-05-26T04:00:07-04:00 2020-06-30T18:12:17-04:00 We showcase examples of doing deep learning (DL) in the browser - for building explorable explanations to aid insight, for building model inference applications and even, for rapid prototyping and training ML model - using the emerging client-side Javascript libraries for DL. Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/445714 2018-05-26T03:45:07-04:00 2020-06-30T18:12:17-04:00 Visualisation for data science requires an interactive visualisation setup which works at scale. In this talk, we will explore the key architectural design considerations for such a system and illustrate using examples the four key tradeoffs in this design space - rendering for data scale, computation for interaction speed, adaptive to data complexity and responsive to data velocity. Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/432660 2018-03-09T22:21:49-05:00 2020-06-30T18:12:17-04:00 A practioners' perspective on doing Deep Learning for Image Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/432451 2018-03-08T14:22:25-05:00 2020-06-30T18:12:17-04:00 Building and scaling data science capability is an imperative for enterprises and startups aiming to adopt a data-driven lens for their business. However, crafting a successful data-science strategy is not straightforward and requires answering the following questions: - Strategy & Tactics: What part of the business should I target first for adoption? Should I take a jump-start approach or a bootstrap approach? - Process & Systems: How should I set up an initial process for data science? How to integrate data-driven processes with existing business systems? - Structure & Roles: Should I adopt a functional or a business-focused data science structure? What specialized roles should I be hiring for Data engineering, ML expert, Visualisation expert, and /or Data Analyst? - Skills & Competencies: How do I up-skill and build differentiated data-science competency across the organization? - Tools & Stack: Should I build a vertical or horizontal data science stack? How do I integrate data science models with existing applications? - Engineering & Technical: What are the pitfalls to watch out for? How to avoid pre-mature over-engineering of data science? How to manage the ongoing technical debt for data science? Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/401445 2017-07-29T02:23:14-04:00 2020-06-30T18:12:17-04:00 What if you could write an interactive data visualisation in the same easy declarative way that you can write markdown to create HTML pages. In this talk, I explain how using a declarative grammar based approach can dramatically speed up the creation of interactive visualisation. I will also talk about a small tool called Visdown, which is open source and created using the excellent vega data visualisation library. You only need to learn the grammar and principles of interactive graphics, and you can then start your own journey in crafting interactive dashboards. Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/389039 2017-04-26T03:14:08-04:00 2020-06-30T18:12:17-04:00 Business data dashboards have gone through renaissance in the last five years with enhanced interactivity and easier tools to create them. However, most dashboard designed today are either at the 'pretty KPI decoration' end or at the 'detailed data explorers' end of the spectrum. They both fail to provide the level of insight communication, that we all see from purposeful designed visualisations in data journalism. This talks aims to understand the missing ingredients in current business dashboard design and aims to start a discussion on how we could rethink data visualisation for them. In the talk, I will explore a few guiding principles that could help us do this. 1. Amplify Cognition 2. Think Datum First 3. Show Single & All 4. Visualise Uncertainty 5. Towards Compositions 6. First Class Annotation 7. Layer Interactions Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/381774 2017-03-02T19:25:15-05:00 2020-06-30T18:12:17-04:00 A thoughtful approach for building your Data Science skills and portfolio. Amit Kapoor (@amitkaps) tag:speakerdeck.com,2005:Talk/380352 2017-02-20T22:41:07-05:00 2025-06-06T22:53:37-04:00 Crafting a talk that engages, inspires and connects with the audience requires time and effort. Plan for it. Here is a short guide to do so, in four sections ― See the Idea, Show the Visual, Tell the Story, and Engage the Audience. Read this guide and then start working on your talk. See the web version at http://amitkaps.com/stories and you can buy a pdf version, if you want from there. Amit Kapoor (@amitkaps) 2020-06-30T12:55:37-04:00