Schedules

Event Schedule

We are in the process of finalizing the sessions. Expect more than 40 talks at the summit. Please check back this page again.
Below is the schedule from MLDS 2019

Expand All +
  • Day 1

    January 30, 2019

  • "We live in a world dominated by anytime, anywhere computing leading to unprecedented growth in data. A dominant trend and a reality that has engulfed the industry is ‘Artificial Intelligence’ and its impact in deriving insights out of such massive corpuses of data. At this juncture, Industry is definitely at the cusp of innovation intended to last a decade or two. AI as a stream stands tall in taking on the challenge to solve some of the dominant scenarios in the industry. As a Data Scientist, you need an innovative platform to rendezvous with Data, and carry out experiments in realizing your Machine Learning Models. More than ever there is a need for us to take a holistic view of a platform that operates with the base premise of democratizing AI. Welcome to the world of Azure Machine Learning Services. Join this session as we explore the possibilities of Machine Learning on Azure in enabling you to infuse intelligence into your solutions. "
    Keynotes

  • In today's Analytics world, Augmented Analytics revolutionizing the way users can find Insights from complex data. Predictive and ML tasks use to be data science game. This session will address how Augmented Analytics helps to fill the skill gaps and bring Predictive and ML to wide spectrum of users ranging from Data Scientist, Citizen Data Scientist to End Business users. You have an opportunity to explore and Hands-on with latest SAP Analytics solution in Augmented Analytics space. This session comprises of both lecture and Hands-on, presented by SAP Experts from Analytics Product Management. Ø Talk on Augmented Analytics - Ashokkumar KN, Sr Product Manager SAP Analytics. Ø Augmented Analytics in Action - Treasure hunt, finding Hidden insights. - Ashutosh Rastogi, Product Manager SAP Analytics - Karthik Kanniyappan, Product Manager SAP Analytics. Prerequisites: Participants to bring Laptop with internet connection."
    Masterclasses

  • Lots of good engineering practices (automation, obsessive focus on productivity, etc) are a vital part of building a functional data science org.
    Tech Talks

  • The aim of this talk is to discuss the potential of Bayesian networks in the area of epidemic detection and description. Bayesian network is a type of Probabilistic graphical models used to capture domain knowledge in cause-effect relationships. The model can be learned on big epidemic data collected in hospitals explaining the causal structure present in the data set. The learned model can then be used to timely predict the outbreak and explain the cause of the spread.
    Tech Talks

  • Machine learning is now a reality. We need to democratize data and machine learning as a tool inside organizations and enable developers to tap into its potential in everything we can do. In this session we will see various ways we can democratize AI/ML inside your team and organization.
    Keynotes

  • The aim of this talk is to demonstrate a solution that empowers the customers to make data driven business decisions. Various industries like, Clinical, Logistics, Automotive etc., are facing real challenge analyzing voluminous, unstructured, and disparate data. The problems are mainly of two types – a. To determine and prevent probable issues and b. To provide a complete solution to a problem that has already taken place. By using Oracle’s Autonomous technology, Customer can seamlessly integrate various sources and store it like, Data Lake and visualize and predict insights on data. Oracle’s Cloud provides the best environment to host an enterprise level solution like the one that this one requires.
    Tech Talks

  • One of the biggest challenge any Data Scientist face is to develop Machine Learning algorithms that deliver the same level of performance in a production environment that it delivers in a controlled environment. Many algorithms suffer a significant backlash when they are exposed to a high number of requests that are typical in many online businesses. Furthermore, tracking of relevant metrics is important to judge the success of the model. In this talk, we shall cover some of these nuances and shall discuss approaches that allow us to delivers the optimal solution for the production environment.
    Keynotes

  • Data lakes are emerging as the most common architecture built in data-driven organizations today. A data lake enables you to store unstructured, semi-structured, or fully-structured raw data as well as processed data for different types of analytics—from dashboards and visualizations to big data processing, real-time analytics, and machine learning. Well-designed data lakes ensure that organizations get the most business value from their data assets. In this session, you learn about the common challenges and patterns for designing an effective data lake on the AWS Cloud, with wisdom distilled from various customer implementations.
    Tech Talks

  • Like Human Learning, Machine Learning is a continuous process. It's a journey and a continuous feedback loop that keeps improving itself. While the ever-mysterious human brain can run these cycles effortlessly, to emulate the same in a business environment can be a challenge. Different set of tools to curate data, test ML Algorithms, deploying the run-time at large scale and then consume the analytics can be daunting task and jugglery of tools and deployment scripts. Enter Azure Databricks. Join this session to learn more about this intuitive collaborative workspace that allows Data Engineer, Data Scientists and Data Analysts to work together from ETL to Model Training to Deployment without compromising on security and scale. We will also look at popular use cases and architectures with Azure Databricks.
    Keynotes

  • Topic 1: Evolution of Data in AI - 30 mins ( Shweta Gupta) The session will outline changing landscape of data in our systems, and how this changing landscape is influenced by our need to infuse AI into systems and processes. It will also briefly touch upon the solutions and platform available here to meet the changing requirements of processing data to build AI solutions. Topic 2: Microsft AI ( 30 mins) ( Gurusubramanian Balasubramanian) Engage and learn how Azure’s complete AI stack enables data engineers and data scientists right from data preparation to running ML models at scale leveraging CPU/GPU to get best performance and fast around time while executing complex ML algorithms Topic 3: Use-case Walkthrough:Infusing AI in a real world scenario ( 45 mins) -( Shweta Gupta) In this walkthrough we will pick up a real world scenario in consumer retail, and transform it into a next generation process by infusing AI, advanced data processing and for core pillars of digital transformation. Topic 4 : Ideation workshop ( 70 mins), Sandeep Alur , Shweta Gupta, Guru, Nalin and Kishore In this workshop, we will walk you through the architecture and help you use Azure AI tools for creating adavnaced AI soulutions.
    Masterclasses

  • Walmart stores across the globe offer millions of items procured from thousands of sellers. Sourcing for each market (US, CA, MX, SAMS, UK,…) is being carried out independently. There can be huge economies of scale in joint global sourcing of items for sale across markets. In order to aggregate each market’s data into a Walmart International perspective, there is a need to generate a global item taxonomy hierarchy leveraging the item categorization of the US, based on their attributes. This will help group similar items/ finelines sold across different markets with the aim of driving EDLC through better negotiations with sellers. It will also enable all the International Markets to operate cohesively, since common item taxonomy can be leveraged across the retail domains.
    Tech Talks

  • Pareto doesn’t apply to ML. 80 % of the time needs to be spent doing things that give 80% of value. Data cleansing and scaling ML. Data cleansing takes half of the 80% of the time:) and the rest is spent in scaling. We will talk about how data cleansing can be done more efficiently with ML, and the focus will turn to what happens next - the need for method, and streamlined, thought-through processes that can enable organizations to go quickly and reliably from raw data to features, whether they are used for training models, or as input into models that are in production. It is this kind of thinking that will enable organizations cut down their time-to-market with every subsequent new model or version of existing model. Today, training and testing models is not optional. Models in production are becoming mission critical, so it’s crucial that you have a thought-through approach as to your stack, your process, your discipline so that you can evolve your pipelines and models quickly and reliably as the usecases proliferate or change. Some of the areas I will touch upon in this talk are: Data Quality and Validation Reproducibility and Auditability Documentation and Discoverability Data Scientist Enablement and Guardrails
    Tech Talks

  • India is going through a major regulatory changes and thus it throws two way challenge to us i.e. to design a compliant system and second to design a "build model" of Analytics / ML system. Lets learn how FAANG companies do that.
    Keynotes

  • With the increase of computing power and availability of data - supervised algorithms are solving efficiently most of classification problems in text. However problems like entity recognition still remains to be an area of craft and domain knowledge. For instance in a chatbot - the 'what' part is easily recognized with the right ML model and right amount of data. The 'Who' of it is still a challenge. In the talk - we will take a deeper look at named entity recognition, the stack of techniques, their applications, strategies and real-life challenges while applying those.
    Tech Talks

  • Financial Companies are rapidly responding to changing technology landscape. In this session, we will discuss how Capital One is preparing itself for the future and is leading the transformation to an AI first company. We will focus on a specific use case of applying OCR to automate loan funding processes. We will discuss latest approaches and why they work, their applications and caveats. The approach learnt here can be applied to other similar use cases as well (e.g. handwriting recognition, audio analysis). Attendees will gain a good appreciation of the methodology and application to their own business case post this session. The session assumes a basic understanding of CNN and LSTM, although they would be briefly covered in the talk.
    Keynotes

  • For a manufacturing heavy, distributor driven, B2B business, it is important to supply the product on time in a cost effective manner. We improved the demand sensing of the end consumers by better demand forecasting and synced the production planning. Along with that, fine-tuned the supply network design to better manage the delivery and cost. This holistic solution is a demand driven supply network.
    Tech Talks

  • This session will focus on how text can be represented for ML tasks and how one can discover "aspects" (key terms) from a given text and then find out the sentiment associated with each aspect term This technique can be used, for example to analyze travel reviews and find out how people feel about a given travel destination. Do travelers like food more or the ambience or the hospitality of the local population. Pre-Requesites: Basic knowledge of python. Ensure you have a laptop with admin privileges and following components installed: 1. Python Anaconda (https://www.anaconda.com/download/) 2. Spacy (https://spacy.io/) 3. Nltk (https://www.nltk.org/) 4. Gensim (https://radimrehurek.com/gensim/) 5. textblob (https://textblob.readthedocs.io/en/dev/)
    Masterclasses

  • Feature selection in large diimensional spaces is a staple of machine learning. While these can be done in several ways (using covariance matrices, mutual inforrmation, SVD, auto-encoders, MCMCs to name a few methods), we propose using genetic algorithms with a multi-dimensional covariance map to solve the feature selection problem. Because we actively trace and reward benefecial parametric co-occurrences using a Bayesian type conditional cross-over and mutation, we are able to significantly reduce the number of generations required to convergence. I will demonstrate the use of this algorithm in conjunction with SVMs to solve two problems in astronomy.
    Tech Talks

  • Speech is an important part of tomorrow’s computing interface. Most of us communicate far more by talking than typing. However, consumers of Alexa, Siri and Google Now are not using it beyond a narrow set of use cases — playing music, checking for weather/traffic etc. There are hardly any conversations that we are used to the real world. One fundamental issue is that humans are programmed to converse with a speaker or a mobile phone. In this talk we are going to introduce the concept of multimodal conversations — where robots use vision, context understanding from the environment, gesture recognition, emotion delivery through body language and finally a human fallback — that produces more engaging conversations in real world applications.
    Keynotes

  • Day 2

    January 31, 2019

  • Ashish will share his own experience in solving real life challenges using Machine Learning and mention various other industrial use cases that have utilized ML techniques to realize enormous value to the business.
    Keynotes

  • The session will cover an introduction to Azure Machine Learning Services for the newcomer, with demos on various features like Intelligent Data Prep, Automated ML and Hyperdrive Hyperparameter tuning. • Session Agenda: • Introduction to Azure Machine Learning • Intelligent Data Prep • AML Automated ML • Hyperdrive Hyperparameter tuning
    Tech Talks

  • In the last few years, interest in AIML has exploded and spilled over to almost all industries and domains. There is hardly any industry which is deemed safe from disruption by AI. Recent advances in AI/ ML/ DL have pushed the boundaries on what was considered possible and newer product and services which were unthinkable few years ago. This has created huge demand for so called AI/ ML/ DL product and services. But is AI for everyone? Can it be made available to everyone? In our master class, we will take a look at what ingredients are needed to put the power of AI in hands of everyone. In particular, we will look at the open source tools and techniques which can be accessed by anyone aspiring to become data scientist/ AIML professional. In the session we will discuss the open source alternatives across the entire project life cycle. We will also discuss how the techniques have evolved over time and use cases for key techniques. Lastly, we will also look at the road blocks which may prevent successful democratization of AI.
    Masterclasses

  • Does AI only solve big problems of the world or is there something in it for me too? Is AI a thing of the future? Are there jobs in AI for me? Will AI take away my job? Shouldn’t I be scared of the humanoids? How trustworthy is AI? These and many more are some burning questions in the mind of a developer today. AI for You brings together thought leaders from different domain to discuss key themes around AI from a developer’s perspective. Join the discussion, get your questions answered by the experts, and get ready for a career in AI.
    Keynotes

  • As technologies like AI, the IoT, advanced analytics, and blockchain become more mainstream, businesses must adapt or risk getting left behind. Responding to individual customer needs, engaging talent in new ways, and creating disruptive business models are critical business imperatives. By becoming an Intelligent Enterprise, you can achieve these goals – and more. Join team SAP to experience how Intelligent enterprises effectively use their data assets to achieve their desired outcomes faster – and with less risk.
    Keynotes

  • In any Business model Customer is king. And King deserves best services in less time. Most of our customers are always confused regarding services or product which they like to buy. Ms. Dyuti Lal will take through different aspect of Recommendation and how to build and implement in your business model. In Machine learning Algorithms Recommendation system are classified into two categories – content based and collaborative filtering methods. Modern recommenders approach both of these categories. In tech talk we will go in deeper and cover various aspect of it and challenges faced during implementation.
    Tech Talks

  • How an economic observation from more than a century ago can help data scientists rethink their approach to data science and develop a fresh perspective to both problem-solving and the process of problem-solving? How does this observation play out today across industries and functions? Additionally, the talk will also include practical tips for data scientists on how to benefit from this perspective.
    Keynotes

  • Food when looked from a scientific perspective, where each ingredient is a makeup of organic molecules which are classified into flavor (taste & smell) and nutrition. The process of cooking itself is a chemical and physical transformation of these ingredients. There are 250K+ ingredients and 26K+ flavor compounds. The flavor network amongst these ingredients which is based on the flavor/taste relationship has immense applications in recipe generation, cuisine fusions, recommendation engine, regional taste preferences, etc. We will explore how AI techniques combined with Food Science can lead to the development of novel applications. Organising large volumes from heterogeneous data sources and building a semantic knowledge system can enable the discovery of competitive intelligence using machine learning. Such a one-stop-knowledge-shop allows decision makers in the FMCG and Foodservice industry to watch trends across the food industry news and blogs, scientific research papers, social media, government regulatory reports, competitors and geo-spatial, demographic and psychographic surveys. Seed trends in the ontology can further power an artificial intelligence system to recognize emerging, waning and stable trends to inform FMCG product and recipe ideation. Trends can be used to identify new recipe fusions which will be used for menu innovation by including various of hyper-local taste, health, fitness, beauty, allergen, dietary preferences, seasonality and many others.
    Tech Talks

  • In this session we will demonstrate some of the great new capabilities that are at offer using cutting edge Oracle Autonomous Technologies like Oracle Autonomous Datawarehouse Cloud and Oracle Analytics Cloud. Hear how Oracle’s application development, integration, systems management, and security solutions leverage artificial intelligence to drive cost savings and operational efficiency for hybrid and multicloud ecosystems. When it comes to forecasting accuracy, Machine Learning often outperforms the traditional models such as ARIMA. In this session we will demonstrate a step-by-step approach on how to use the power of machine learning for forecasting. This session is also a good way to get a sneak peek on working with Machine Learning in Autonomous Datawarehouse Cloud in general. The principles you learn here can be applied to many more machine learning use cases. We will also cover any required ML theory, so no previous knowledge of ML is required to attend this session. After attending this session one will have the basic ingredients to apply ML to their own business cases with Oracle Autonomous Datawarehouse Cloud.
    Keynotes

  • In this masterclass, 1. We discuss the best practises from the TEG Republic Day hackathon, 2. Provide key pivotal tips and tricks that are invaluable in cracking data science competitions 3. Provide hands-on learning on market share predictions for products with multiple attributes, using Medicare enrolments as a case study.
    Masterclasses

  • Deep Learning (DL) researchers use GPUs and CPUs to training their models. The target deployment environment introduces various challenges that are typically not present in the training environment. How to increase the performance of trained DL model during inference phase? Dr. Sunil shall present various options for improving performance at Inference stage. He describes Intel’s inference engine as one option to run DL models in Caffe/Tensorflow/Keras using CPUs. He discusses implementation for two deep learning based healthcare models and share results on the performance benchmarks. He highlights how the CPUs/USB stick can be used to its maximum potential and still not compromising on the quality of output at exceptional speeds.
    Tech Talks

  • As part of this topic will cover the following details Planning a scalable lifecycle for analytical models (POC to Pilot to Production) Planning for scalable software layer (R vs Python vs Scala; MapReduce) Planning for scalable hardware layer (achieving the cost vs performance trade-offs) Planning for scalable Devops (kubernetes based deployment path) Checklist for success in deploying Analytics models at scale (dos and donts)
    Tech Talks

  • The expected growth in AI business is simply not happening at the anticipated rate. Consequently, if someone is an AI professional, there’re not many big projects that are coming their way. I’d like to talk about why this is the case, and if you’re an AI professional, how to prepare.” The first challenge is lack of skillset in the IT employee base. Are employers upskilling their employees? Are they investing enough? AI requires specialized knowledge: statistics, algorithms and computer science, not civil or mechanical engineering. So, well-intentioned employees starting on the AI journey get stuck. Additionally, there is only a small talent pool. 500,000 jobs are available in AI, but only 50,000 computer science grads per year, out of which only 20-30% are of some quality. Second, companies need to pick processes carefully that they want to incorporate AI into. If not, maybe RPA or analytics is a better fit. For example, for routine invoice processing, payroll processing, etc., RPA can fit the bill nicely. Third, the lack of infrastructure sets us back as a country in terms of AI adoption. Basic telephone access and WiFi are missing; how can we talk about drones and video processing in AI? If you’re an AI professional, the key is to learn the specific skills, and encourage your company to select incoming employees from the computer science, applied mathematics or statistics background. Second, help your business leaders select the correct business processes for AI enablement. Finally, learn to work with the reality of low-quality infrastructure, and pick use cases that have a high chance of success even with the inferior infrastructure.
    Keynotes

  • In the current competitive market scenario, using only linear models to solve predictive modeling problems will leave us behind. The trend is moving towards using the machine learning and deep learning algorithms in analytics industry to solve predictive modeling problems because of their low bias, higher predictive power and dependencies on more features. The real challenge is to explain these black box algorithms. In this talk, I will discuss different ways of explaining machine learning models.
    Keynotes

  • As streaming media is going digital and increasing its foothold as an essential ingredient in everyone's life, we are fortunate that technological developments are simultaneously powering machine learning capabilities into the ecosystem. Even conservatively thinking, the next five to ten years will completely alter the way entertainment is consumed. And that is best reflected in the way disruptions are being witnessed in the business models around content creation and content consumption. To go a step further, I'd structure the talk in the following components - What AI techniques are disrupting the video streaming business - Image and video recognition, recommendation engines, media databases, metadata generation, predictive content delivery, language translation/ captioning, predictive marketing, programmatic ad buying, connected devices, virtual assistants, AR and VR Real-life example from Voot - AI applications in practice and plans, including plans for intelligent adaptive bitrate streaming, self-reliant marketing automation plans and predicting catastrophic failures; methods and outcomes How AI movement will transform business models - New content formats, new ad formats, new streaming protocols, new gaming formats, new interaction touch points, personalisation, AR, VR and blockchain based monetisation models, connected devices, economics of content creation and delivery and future business models
    Keynotes

Extraordinary Speakers

Meet the best Machine Learning Practitioners & Researchers from the country.

  • Early Bird Pass

    Available from 1st Nov to 4th Dec 2019
  • Access 1-3 Stages
  • Access 50+ Speakers
  • Access to Exhibition Area
  • Access to food & beverage
  • All access, 2-day pass
  • Group discount available
  • 4,500 + taxes
  • Regular Pass

    AVAILABLE FROM 4th DEC 2018 to 8th Jan 2020
  • Access 1-3 Stages
  • Access 50+ Speakers
  • Access to Exhibition Area
  • Access to food & beverage
  • All access, 2-day pass
  • Group discount available
  • 6,000 + taxes