You have been temporarily restricted. Please be more thoughtful when adding content for your portfolio. Your portfolio and digital media kit and should be reflective of the professional image you wish to convey. Accounts may be temporarily restricted if we receive reports of spamming or if the system detects excessive entries.
Membership
Publish your original ideas on the Thinkers360 platform!
This feature is available for Pro and Pro-Plus Members Only.
Speaker Bureau functionality whereby individuals can be featured speakers within our Speaker Bureau service and enterprises can find and work with speakers.
This feature is available for Pro, Pro-Plus, Premium and Enterprise Members Only.
Highlight your featured products and services within our company directory for enhanced visibility to active B2B buyers worldwide. This feature is available for Pro, Pro Plus, Premium and Enterprise Members Only.
Contribute to the Thinkers360 Member Blog and have your thought leadership featured on our web site, newsletter and social channels. Reach our opt-in B2B thought leader community and influencer marketplace with over 100M followers on social media combined!
You’ve reached your daily limit for entering quotes. Please only add personally-authored content which is reflective of your digital media kit and thought leadership portfolio.
Thinkers360 Content Library
For full access to the Thinkers360 content library, please join ourContent Planor become a contributor by posting your own personally-authored content into the system viaAdd PublicationorImport Publication.
Dashboard
Unlock your personalized dashboard including metrics for your member blogs and press releases as well as all the features and benefits of our member plans!
Interested in getting your own thought leader profile? Get Started Today.
Debmalya Biswas
Director, Data Analytics & AI at Wipro
Lausanne, Switzerland
Debmalya Biswas has over 20 years of experience working on AI/ML, Big Data Analytics, Privacy & Security related topics; most recently as Principal AI Architect at Philip Morris International (PMI) where he led PMI's AI Center of Excellence.
He has previously worked for leading technology companies, e.g., Nokia Research, SAP Research, Oracle, Swisscom, and the Swiss startup Iprova.
He has authored 50+ patents, 30+ scientific articles, is a regular speaker at leading AI/ML conferences and is a member of the AI4EU Evaluation Committee. He holds a PhD in Computer Science from INRIA, France.
Available For: Advising, Authoring, Consulting, Influencing, Speaking Travels From: Switzerland Speaking Topics: Artificial Intelligence, Data Science, Privacy
Debmalya Biswas
Points
Academic
45
Author
9
Influencer
15
Speaker
16
Entrepreneur
1160
Total
1245
Points based upon Thinkers360 patent-pending algorithm.
Hands-on: Optimizing and benchmarking Body Pose Estimation models
Medium.com
September 07, 2021
Abstract. Nvidia recently announced the availability of 2D Body Pose Estimation Model as part of the Transfer Learning Toolkit 3.0. In this article, we provide a detailed tutorial to train and optimize the model. We further provide benchmarking results with respect to another widely used open source model for perception tasks: OpenPifPaf—allowing you to decide which model to use when.
Edge AI Framework for Healthcare Applications
4th IJCAI Workshop on AI for Ageing, Rehabilitation and Intelligent Assisted Living (ARIAL)
August 09, 2021
Edge AI enables intelligent solutions to be deployed on edge devices, reducing latency, and providing strong privacy guarantees. This makes Edge AI based applications very promising for a range of Healthcare solutions. Unfortunately, achieving efficient and accurate execution of AI/ML algorithms on edge devices, with limited power and computational resources, raises several deployment challenges. To address these challenges, we present a platform agnostic framework to (1) foster model-to-platform adaptations, (2) enable validation of the deployed models, and (3) empower engineers to do it efficiently using repeated, but rapid development cycles. We show the practical utility of the proposal by applying it on a real-life Healthcare Body-pose Estimation App. The app monitors patients in their beds and signals an alarm when someone falls off their bed, has an accident, or are in general experiencing restricted movement. This can be a big problem for residents of hospitals and old-age homes. We describe the full deployment pipeline including selection of the object detection algorithms for face-landmarks and body pose detection, e.g., RetinaFace and OpenPifPaf; and porting them to the Nvidia Jetson NX platform.
Privacy Preserving Chatbot Conversations
3rd IEEE International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)
December 31, 2020
With chatbots gaining traction and their adoption growing in different verticals, e.g. Health, Banking, Dating; and users sharing more and more private information with chatbots — studies have started to highlight the privacy risks of chatbots. In this paper, we propose two privacy-preserving approaches for chatbot conversations. The first approach applies ‘entity’ based privacy filtering and transformation, and can be applied directly on the app (client) side. It however requires knowledge of the chatbot design to be enabled. We present a second scheme based on Searchable Encryption that is able to preserve user chat privacy, without requiring any knowledge of the chatbot design. Finally, we present some experimental results based on a real-life employee Help Desk chatbot that validates both the need and feasibility of the proposed approaches.
Delayed Rewards in the Context of Reinforcement Learning based Recommender Systems
AAI4H — Advances in Artificial Intelligence for Healthcare Workshop, co-located with the 24th European Conference on Artificial Intelligence (ECAI 202
September 30, 2020
We present a Reinforcement Learning (RL) based approach to implement Recommender systems. The results are based on a real-life Wellness app that is able to provide personalized health related content to users in an interactive fashion. Unfortunately, current recommender systems are unable to adapt to continuously evolving features, e.g. user sentiment, and scenarios where the RL reward needs to be computed based on multiple and unreliable feedback channels (e.g., sensors, wearables). To overcome this, we propose three constructs: (i) weighted feedback channels, (ii) delayed rewards, and (iii) rewards boosting, which we believe are essential for RL to be used in Recommender Systems. Finally, we also provide some implementation details on how the Wellness App based on Azure Personalizer was extended to accommodate the above RL constructs.
Reinforcement Learning based HVAC Optimization in Factories
11th ACM International Conference on Future Energy Systems (e-Energy)
June 30, 2020
Heating, Ventilation and Air Conditioning (HVAC) units are responsible for maintaining the temperature and humidity settings in a building. Studies have shown that HVAC accounts for almost 50% energy consumption in a building and 10% of global electricity usage. HVAC optimization thus has the potential to contribute significantly towards our sustainability goals, reducing energy consumption and CO2 emissions. In this work, we explore ways to optimize the HVAC controls in factories. Unfortunately, this is a complex problem as it requires computing an optimal state considering multiple variable factors, e.g. the occupancy, manufacturing schedule, temperature requirements of operating machines, air flow dynamics within the building, external weather conditions, energy savings, etc. We present a Reinforcement Learning (RL) based energy optimization model that has been applied in our factories. We show that RL is a good fit as it is able to learn and adapt to multi-parameterized system dynamics in real-time. It provides around 25% energy savings on top of the previously used Proportional-Integral-Derivative (PID) controllers.
Self-improving Chatbots based on Reinforcement Learning
4th Multidisciplinary Conference on Reinforcement Learning and Decision Making (RLDM)
July 31, 2019
We present a Reinforcement Learning (RL) model for self-improving chatbots, specifically targeting FAQ-type chatbots. The model is not aimed at building a dialog system from scratch, but to leverage data from user conversations to improve chatbot performance. At the core of our approach is a score model, which is trained to score chatbot utterance-response tuples based on user feedback. The scores predicted by this model are used as rewards for the RL agent. Policy learning takes place offline, thanks to an user simulator which is fed with utterances from the FAQ-database. Policy learning is implemented using a Deep Q-Network (DQN) agent with epsilon-greedy exploration, which is tailored to effectively include fallback answers for out-of-scope questions. The potential of our approach is shown on a small case extracted from an enterprise chatbot. It shows an increase in performance from an initial 50% success rate to 75% in 20–30 training epochs.
Privacy preserving minimal observability for composite transactional services
Springer Discrete Event Dynamic Systems Journal
January 23, 2014
For complex services composed of many (component) services, logging is an integral middleware aspect, especially for providing transactions and monitoring. In the event of a failure, the log allows us to deduce the cause of failure (diagnosis) and recover by compensating the executed services (atomicity). However, for heterogeneous services with parts of the functionality provided by multiple organizations, logging details of all executed services is often impracticable due to privacy/security constraints. Also, logging is expensive in terms of both time and space. Thus, we are interested in determining the minimal number of services that need to be logged, and which is still sufficient to know with certainty the actual sequence of executed services from any given log. Further to privacy issues, the complexity of determining a minimal set of such services to log is actually NP-Complete. To solve both issues, we resort to considering each component service as a grey box. Logs are recorded and kept local to each component, and a black-box view of the implementation details of each component is provided. In particular, a service which is reused as a component several times (often observed in real-life services) need not be re-computed each time. We show that this dramatically decreases the complexity up to 2 exponentials. For large monolithic component services that cannot be decomposed simply, we also provide heuristics to compute a small (but not necessarily minimal) number of services to log, and experimentally analyze their accuracy and performance.
Privacy preserving and transactional advertising for mobile services
Springer Computing Journal
June 15, 2013
The many sensors embedded in phones nowadays provide advanced sensing capabilities that make it possible to capture real-time information about the user and his surroundings. There are already examples of apps/services that use this information to provide highly useful and contextual advertisements to the users. However, users are still reluctant to share their personal data with advertisers due to their privacy implications (if misused). In this work, we provide protocols that allow users to store their sensor data on third party (untrusted) cloud servers. The data is stored in encrypted form, hence protected from the cloud provider. The advertisements are also stored on the server. They are customized to potential users. The server selects the advertisements appropriate to the users, based on their sensor values, and forwards them to the users. We consider two cases: (i) appropriate advertisements are sent to individual users automatically; and (ii) advertisements are sent to groups of users after getting permission from the group members. In both cases, the concurrency control protocols performed by the cloud provider ensure that the data and advertisements are ‘fresh and consistent’. This is to avoid situations where served advertisements are not in sync with the user’s current context or the advertisements have already expired. The above is achieved by integrating transactional and cryptographic primitives, such as atomic uploads, optimistic concurrency control, searchable encryption and homomorphic encryption. Finally, experimental results are given to illustrate the practical feasibility and scalability of the proposed protocols.
Distributed and minimal usage control
Springer Service Oriented Computing and Applications
May 26, 2012
With more and more personal data being collected and stored by service providers, there is an increasing need to ensure that their usage is compliant with privacy regulations and user preferences. We consider the specific scenario where promised usage is specified as metric temporal logic policies, and these policies can be verified against the database usage logs. Given the vast amount of data being collected, scalability is very important. In this work, we show how such usage monitoring can be performed in a distributed fashion for an expressive subset of policies. Experimental results are given for a real-life use case to show the genericness and scalability of the results.
Formalizing visibility characteristics in hierarchical systems
Elsevier Data & Knowledge Engineering Journal
August 01, 2009
We consider hierarchical systems where nodes represent entities and edges represent binary relationships among them. An example is a hierarchical composition of Web services where the nodes denote services and the edges represent the parent–child relationship of a service invoking another service. A fundamental issue to address in such systems is, for two nodes X and Y in the hierarchy, whether X can see Y, that is, whether X has visibility over Y. The visibility could be with respect to certain attributes like operational details, execution logs and security related issues. In a general setting, X seeing Y may depend on (i) X wishing to see Y, (ii) Y wishing to be seen by X, and (iii) other nodes not objecting to X seeing Y. In this paper, we develop a generic conceptual model to express visibility. We study two complementary notions: sphere of visibility of a node X that includes all the nodes in the hierarchy that X can see; and sphere of noticeability of X that includes all the nodes that can see X. We also identify dual properties, coherence and correlation, that relate the spheres of different nodes in special ways and also relate the visibility and noticeability notions. We study some variants of coherence and correlation also. These properties give rise to interesting and useful visibility and noticeability assignments, and their representations.
Compositional AI: Fusion of AI ML Services
Data Fusion International Conference
March 31, 2021
Enterprise adoption of AI/ML services has significantly accelerated in the last few years. However, the majority of ML models are still developed with the goal of solving a single task, e.g., prediction, classification. In this talk, Debmalya Biswas will present the emerging paradigm of Compositional AI, also known as, Compositional Learning. Compositional AI envisions seamless composition of existing AI/ML services, to provide a new (composite) AI/ML service, capable of addressing complex multi-domain use-cases. In an enterprise context, this enables reuse, agility, and efficiency in development and maintenance efforts.
The talk will introduce Compositional AI, outline current state-of-the-art and challenges, and provide directions for future exploration to accelerate the enterprise readiness of Compositional AI.
Wireless routing
United States Patent Office
April 27, 2021
According to an example aspect of the present invention, there is provided an apparatus comprising a receiver configured to receive, from a first node, an indication of privacy preference and, from a second node, an indication relating to an accuracy at which the second node can determine a location of the first node, and at least one processing core configured to determine routing information based at least in part on the indication of privacy preference and the indication relating to the accuracy.
Method and apparatus for privacy policy management
European Patent Office
January 15, 2021
Various methods are provided for determining run-time characteristics of an application at the time of installation and/or modification. Based on the determined run time characteristics, various methods control the installation and/or modification of the application based on a user privacy profile. One example method may comprise receiving a request to modify an application. A method may further comprise determining whether a conflict is present between the application and a user privacy profile. A method may further comprise causing the determined conflict and at least one conflict resolution to be displayed in an instance in which a conflict is determined. A method additionally comprises causing a user privacy profile to be modified in an instance in which an indication of acceptance is received in response to the displayed at least one conflict resolution.
Local authentication
European Patent Office
January 22, 2020
In accordance with an example aspect of the present invention, there is provided an apparatus comprising at least one processing core and at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processing core, cause the apparatus at least to obtain a first sensor output from a first sensor, receive from a second apparatus a first received sensor output and compare the first sensor output to the first received sensor output, and responsive to the outputs matching, grant a first level of access to the second apparatus.
Apparatus and associated methods for use in live navigation
United States Patent Office
January 07, 2020
An apparatus comprising a processor and memory including computer program code, the memory and computer program code configured to, with the processor, enable the apparatus at least to: for a defined road journey through a road network from a start location to an end location, assign respective sub-portions of the defined road journey to each of a group of mobile devices associated with one or more travellers for the road journey according to at least one predefined assignment criterion, wherein each of the group of mobile devices is configured for provision of live mapping data for live navigation of the respective sub-portion of the road journey.
Computer-enabled method of assisting to generate an innovation
United States Patent Office
October 29, 2019
According to the present invention there is provided a computer-enabled method of assisting to generate an innovation, the method comprising the steps of: retrieving from a database a first set of more than two documents belonging to a first domain (D1); retrieving from said database a second set of more than two documents belonging to a second domain (D2); selecting all possible combinations of documents from the first set with all documents in said second set, and for each combination of documents: determining a composite novelty score, a composite proximity score and a composite impact score; and based on all of the determined composite novelty scores and/or composite proximity scores and/or composite impact scores, providing a recommendation which can assist to generate an innovation.
Privacy Management
United States Patent Office
October 15, 2019
According to an example aspect of the present invention, there is provided an apparatus comprising a memory interface configured to access a plurality of privacy feature requirement sets, each privacy feature requirement set being associated with a source code segment, and at least one processing core configured to select a first source code segment from among the source code segments based on a comparison between a privacy feature requirement set associated with the first source code segment and a current privacy feature requirement set, and to derive a privacy feature deviation set using the current privacy feature requirement set and the selected first source code segment.
Location privacy
United States Patent Office
September 03, 2019
According to an example aspect of the present invention, there is provided an apparatus comprising at least one receiver configured to receive a route indication specifying an initial point and an end point of a route, and at least one processing core configured to determine a plurality of candidate routes based at least in part on the route indication and a privacy preference indication, and to determine for each of the plurality of candidate routes a privacy overhead route.
Method and apparatus for policy adaption based on application policy compliance analysis
United States Patent Office
June 25, 2019
An approach is provided for policy adaption based on application policy compliance analysis. The compliance platform processes and/or facilitates a processing of one or more policy compliance logs associated with at least one application to determine one or more policy compliance profiles associated with the at least one application. The compliance platform determines one or more contexts under which the at least one application operates. The compliance platform causes, at least in part, an association of the one or more policy compliance profiles with the one or more contexts. The compliance platform then processes and/or facilitates a processing of user contextual information, user application use information, or a combination thereof against the association, the one or more policy compliance profiles, the one or more contexts, or a combination thereof to determine one or more adaptions to one or more policies associated with the at least one application.
Method and apparatus for tagged deletion of user online history
United States Patent Office
March 12, 2019
An approach is provided for deleting a user's online data across different services and platforms based on contextual selection criteria. The deletion manager determines at least one request to delete data associated with at least one user, the request specifying at least in part one or more contextual parameters. The deletion manager determines one or more data records associated with the at least one user from one or more services, one or more applications, or a combination thereof. The deletion manager causes, at least in part, a deletion of the one or more data records based, at least in part, on whether the data at least substantially meet the one or more contextual parameters.
Method and apparatus for providing privacy policy generation based on in-game behavior data
United States Patent Office
January 29, 2019
An approach is provided of generating real-world privacy policies based on in-game behavior. A privacy policy platform determines in-game behavior data associated with at least one user while the at least one user is playing at least one location-based game. The privacy platform causes, at least in part, a mapping of the in-game behavior data to one or more game locations within the at least one location-based game. The privacy platform further causes, at least in part, a correlation of the one or more game locations to one or more real-world locations. The privacy platform then causes, at least in part, a generating of one or more privacy policies for the one or more real-world locations based, at least in part, on the in-game behavior data mapped to the correlated one or more game locations.
Method and apparatus for secure advertising
European Patent Office
January 15, 2019
A method, apparatus, and computer program product are provided to serve advertisements based upon the current context of the user in a manner that both ensures the relevancy of the advertisements and maintains the privacy of the contextual information of the user. In this regard, a method, apparatus and computer program product are provided to receive encrypted context data from one or more users including one or more trapdoors associated with one or more keywords and to receive advertisement information from one or more advertisers including a searchable encryption. Further, the method, apparatus and computer program product are provided to cause an advertisement to be provided to a first user based upon the encrypted context data of the first user and the advertisement information from one or more advertisers.
Method and apparatus for updating or validating a geographic record based on crowdsourced location data
United States Patent Office
December 04, 2018
Methods and apparatuses are provided for validating crowdsourced location data. A validation platform causes, at least in part, a determination of location data reported by at least one user, at least one device associated with the at least one user, or a combination thereof. The validation platform processes, and/or facilitates a processing of personal detail information associated with the at least one user, the at least one device, or a combination thereof to cause, at least in part, a validation of the location data.
Location identification
European Patent Office
October 24, 2018
A method comprises receiving an indication of an activity to be performed, obtaining an indication of a current location of a user, transmitting, to a server, information relating to an area in which the location for the activity is to be identified, said information including the current location and one or more criteria relating to activity type and/or required privacy level, receiving, from the server, location information identifying one or more candidate locations for the activity. If a suitable candidate location cannot be identified within the area, further information regarding the candidate locations may be obtained, and comments regarding future conditions or shortcomings of the candidate location may be displayed, or the activity may be divided into two or more sub-tasks to be performed at different locations. The user's progress in performing the activity may be monitored, so that the displayed candidate locations can be updated.
Method and apparatus for determining privacy policy for devices based on brain wave information
United States Patent Office
September 04, 2018
An approach is provided for processing sensor data to determine an identification of at least one object. The approach involves determining brain wave data associated with at least one user during an exposure of the at least one object to at least one user. The approach also involves processing the brain wave information to determine a sensitivity information of the at least one user to the at least one object. The approach further involves causing a configuration of one or more privacy policies of at least one device based, at least in part, on the sensitivity information.
Method and apparatus for providing privacy profile adaptation based on physiological state change
United States Patent Office
April 17, 2018
An approach is provided for adapting privacy profiles to respond to changes in physiological state. The policy platform may process and/or facilitate a processing of sensor information to determine at least one change in one or more physiological states of at least one user, wherein the at least one user is associated with at least one context, at least one activity, or a combination thereof. Then, the policy platform may cause, at least in part, a modification of at least one privacy profile for at least one device associated with the at least one user based, at least in part, on the at least one change in the one or more physiological states, the at least one context, the at least one activity, or a combination thereof, wherein the modification of the at least one privacy profile includes, at least in part, an enabling or a disabling of one or more privacy services operating at least at least one device.
Method and apparatus for providing group route recommendations
United States Patent Office
January 30, 2018
An approach is provided for providing alternative route recommendations for a group of nearby users in a privacy preserving manner. A recommendation platform determines at least one group of one or more users within proximity of at least one location-based service, at least one location-based event, or a combination thereof. A recommendation platform processes and/or facilitates a processing of one or more datasets associated with the group to cause, at least in part, a determination of whether to recommend the at least one location-based service, the at least one location-based event, or a combination thereof to the group. A recommendation platform causes, at least in part, a presentation of the at least one location-based service, the at least one location-based event, or a combination thereof to the group, the one or more users, or a combination thereof based, at least in part, on the determination of whether to recommend.
Method and apparatus for generating privacy ratings for applications
United States Patent Office
August 01, 2017
An approach is provided for generating privacy ratings for applications. A privacy ratings platform determines use information associated with one or more applications executing on one or more devices. By way of example, the use information is determined based, at least in part, on usage data associated with one or more input sources, one or more components, one or more categories of personal information, or a combination thereof associated with the one or more devices. The privacy ratings platform then processes and/or facilitates a processing of the use information to determine one or more privacy ratings for the one or more applications.
Determination of a filtered map interaction descriptor
United States Patent Office
July 18, 2017
A method comprising receiving information indicative of a map operation that is associated with a map object, the map operation being invoked by a user, determining a map interaction descriptor that comprises information indicative of the map object, information indicative of the map operation, and information indicative of a map object property, determining a privacy setting that is associated with the map interaction descriptor based on at least a portion of the map interaction descriptor, determining a filtered map interaction descriptor based on the privacy setting and the map interaction descriptor, such that the filtered map interaction descriptor is absent at least a portion of at least one of the information indicative of the map object, the information indicative of the map operation, or the information indicative of the map object property, and causing sending of information indicative of the filtered map interaction descriptor to a separate apparatus is disclosed.
Method and apparatus for transforming application access and data storage details to privacy policies
United States Patent Office
July 11, 2017
An approach is provided for transforming application access and data storage details associated with a mobile device to application specific privacy policies. A policy platform processes and/or facilitates a processing of an application code to identify one or more application programming interfaces invoked by the application code. The policy platform causes, at least in part, a profiling of one or more invocation characteristics of the one or more application programming interfaces. The policy platform further causes, at least in part, a mapping of the one or more invocation characteristics against one or more policy templates to cause, at least in part, a generation of at least one policy for the application code.
Method and apparatus for providing location privacy
European Patent Office
March 01, 2017
An approach is provided for providing real-time controlled location privacy as the location evolves, and providing a user with alternate routes and applications depending on the level of desired location privacy. A location privacy platform determines at least one location associated with at least one device. The location privacy platform also processes and/or facilitates a processing of contextual information associated with the at least one location, the at least one device, one or more applications associated with the at least one device, or a combination thereof to determine one or more privacy metrics for the one or more applications with respect to the at least one location; wherein the one or more privacy metrics relate, at least in part, to an exposure of user data by the one or more applications at the at least one location.
Sending of map interaction data to a program
United States Patent Office
January 17, 2017
A method comprising identifying a source program to send a target program map interaction data set to a target program, identifying a target program map function set that includes at least one target program map function, identifying a source program map function set that includes at least one source program map function, determining a transfer map function set that is a set intersection of the target program map function set and the source program map function set, identifying a source program map interaction data set, determining a target program map interaction data set to include each source program map interaction data item from the source program map interaction data set that corresponds with at least one target program map function of the target program map function set, and sending the target program map interaction data set to the target program is disclosed.
Method and apparatus providing privacy benchmarking for mobile application development
United States Patent Office
November 15, 2016
A method operates, during development of an application program intended to be run on a mobile user device, to perform a computer assisted analysis of the application program to determine at least one user privacy-related aspect of the application program; and to present the determined at least one user privacy-related aspect. The determined at least one user privacy-related aspect may be presented to a developer of the application program. An apparatus and system for performing the method are also disclosed.
Method and apparatus for providing privacy adaptation based on receiver context
United States Patent Office
October 04, 2016
An approach is provided for determining a communication session established between at least one first device of at least one first user and at least one second device of at least one second user, wherein the at least one first device and the at least one second device are in a collaborative context detection relationship. The privacy platform causes, at least in part, a detection of a presence of at least one third user within a proximity threshold of the at least one first device, the at least one second device, the at least one first user, the at least one second user, or a combination thereof, wherein the detection is performed via the collaborative context detection relationship. The privacy platform also determines one or more modifications to one or more privacy profiles for information exchanged over the communication session based, at least in part, on the detection of the presence of the at least one third user.
Method and apparatus for controlling transmission of data based on gaze interaction
United States Patent Office
August 09, 2016
An approach for controlling transmission of data based on gaze interaction. A data transmission determination platform determines one or more gaze metrics for one or more users. The one or more gaze metrics, relate, at least in part, to a level of interaction of each user with an object. The data transmission determination platform then processes and/or facilitates a processing of the one or more gaze metrics, user preference information, propagation information from one or more entities associated with the object, or a combination thereof to determine data to transmit one or more devices associated with at least a subset of the at least one or more users.
Method and apparatus for providing collaborative privacy policies for a shared device
United States Patent Office
January 05, 2016
An approach is provided for determining one or more user inputs for specifying configurable privacy-related data for at least one shared device, wherein the one or more user inputs are associated with one or more users of the at least one shared device. The approach involves processing and/or facilitating a processing of the one or more user inputs to identify at least one potential privacy conflict resulting from the one or more user inputs. The approach also involves causing, at least in part, an initiation of at least one privacy preserving action based, at least in part, on the at least one potential privacy conflict.
Compositional AI: The Future of Enterprise AI
Towards Data Science
April 01, 2021
"it is time to move beyond model precision to a more holistic framework of Compositional AI, enabling maximal reuse of the deployed AI/ML solutions in the enterprise."
Open Source Enterprise AI/ML Governance
Linux Foundation Open Compliance Summit
December 04, 2020
With the growing adoption of Open Source based AI/ML systems in enterprises, there is a need to ensure that AI/ML applications are responsibly trained and deployed. This effort is complicated by different governmental organizations and regulatory bodies releasing their own guidelines and policies with little to no agreement on the definition of terms, e.g. there are 20+ definitions of ‘fairness’. In this article, we will provide an overview explaining the key components of this ecosystem: Data, Models, Software, Ethics and Vendor Management. We will outline the relevant regulations such that Compliance/Legal teams are better prepared to establish a comprehensive AI Governance framework. Along the way, we will also highlight some of the technical solutions available today that can be used to automate these compliance requirements.
Building an enterprise Natural Language Search Engine with ElasticSearch and Facebook’s DrQA
Berlin Buzzwords Conference
June 30, 2019
Modern search engines leverage Natural Language Processing (NLP) and Machine Learning (ML) to improve relevance of results. In this presentation, we focus on the specific field of ‘Enterprise Search’, whose primary goal is to make domain specific company data and documents readily accessible to employees to improve their productivity and promote collaboration. Indeed, any large organization produces vast amounts of documentation about their specific systems, technologies and processes. The question then is - “How can we speed up search-driven activities and enhance user experience for the enterprise?”
Facebook AI research team has developed and open sourced a tool to answer questions by reading Wikipedia articles, called DrQA (Chen, et al. 2017). DrQA is based on a 2-stage Q&A pipeline: (i) Retriever: retrieve the top-k relevant documents (pages), followed by (ii) Reader: determining the most relevant answer span among the retrieved documents (pages). We applied it on an enterprise use-case to search over machine manuals used by factory operators.
We present an architecture integrating ElasticSearch in the DrQA pipeline, which has been contributed upstream and is now available from the official DrQA github repository (https://github.com/facebookresearch/DrQA). The end result is a very scalable search engine that can be deployed on any document repository in your enterprise containing Microsoft Office docs, ppts, emails, pdf documents, etc. Simply point it to your ElasticSearch index and it will be able to provide ‘very precise’ answers based on your documents, thanks to the pre-trained Deep Learning Q&A model of DrQA. We discuss the learnings along the creation and the limitations of such an engine, e.g. scenarios where it excels by identifying precise answers and how it performs compared to a non-ML approach or a typical keyword based search.
Join Thinkers360 for free! Are you a Reader/Writer, Thought Leader/Influencer (looking to increase your earnings), or an Enterprise User (looking to work with experts)?
Hands-on: Optimizing and benchmarking Body Pose Estimation models
Facial Recognition Inference Pipeline for Sentiment Analysis
Compositional AI: the future of Enterprise AI
Edge AI: Deploying AI/ML on Devices
Business applications of Reinforcement Learning
How Privacy Sensitive is your Mobile App?
Federated Learning — Privacy preserving Machine Learning
Open Source Licensing primer for Enterprise AI/ML
Ethical AI: its implications for Enterprise AI Use-cases and Governance
Edge AI Framework for Healthcare Applications
Privacy Preserving Chatbot Conversations
Reinforcement Learning based HVAC Optimization in Factories
Self-improving Chatbots based on Reinforcement Learning
Open Source Enterprise AI/ML Governance
