Sanjay Dorairaj

San Jose, California, United States Contact Info
2K followers 500+ connections

Join to view profile

About

Passionate, execution-focused, tireless technologist with a solid record of having…

Articles by Sanjay

See all articles

Activity

Join now to see all activity

Experience & Education

  • Amazon Web Services (AWS)

View Sanjay’s full experience

See their title, tenure and more.

or

By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.

Licenses & Certifications

Volunteer Experience

  • Save the Children International Graphic

    Member of Tech and Data Impact Council

    Save the Children International

    - Present 1 year 7 months

    Children

  • Our Lady of Refuge Graphic

    Member of Pastoral Council

    Our Lady of Refuge

    - Present 1 year 5 months

    Social Services

    Work through other members of the Pastoral Council to organize and execute activities within the Our Lady of Refuge Parish, San Jose

  • Summit Public Schools (Public Charter Network) Graphic

    Instructor

    Summit Public Schools (Public Charter Network)

    - 1 year 8 months

    Science and Technology

    Teach Python programming

  • Member of the Pastoral Council

    St Francis of Assisi, San Jose

    - 1 year 9 months

    Social Services

    Assisting the Parish with decisions related to the overall functioning of the Parish, especially as it relates to the governing of the various Parish lay ministries.

  • Society of St. Vincent de Paul Graphic

    Volunteer

    Society of St. Vincent de Paul

    - 5 years 1 month

    Social Services

    House visit to clients of St Vincent De Paul to work with them to understand their needs and their current financial situation to see how the SVDP society can appropriately tend to their needs.

Publications

  • TLDR; Abstractive Text Summarization Using Sequence to Sequence Models and an N-Gram Language Model

    See publication

  • Hidden Markov Models Simplified

    Medium

    A simple explanation of Hidden Markov Models

    See publication

  • Simplifying Field Operations Using Machine Learning

    NCTA Technical Papers

    This paper is about machine learning - its definition and its applications. It especially examines the relevance of machine learning from the perspective of the cable’s multiple system operators (MSOs). While there have been some attempts in technical and trade literature to pinpoint the benefits of machine learning to cable service operators, there has not yet been a holistic treatment of the subject, to our knowledge. This paper is an attempt to fill that gap.

    See publication

  • Perspectives on IoT

    SCTE (scte.org)

    Visualizing the possibilities and the promise of IoT, while exciting, is also quite overwhelming. This is because the possibilities are quite frankly, endless. The challenge then is to sort through the possibilities and the associated technological and business challenges and identify opportunities that make sense for an MSO to pursue. This paper provides such a framework for pursuing opportunities with IoT. The framework that is proposed is focused on trying to build lasting value for the MSO…

    Visualizing the possibilities and the promise of IoT, while exciting, is also quite overwhelming. This is because the possibilities are quite frankly, endless. The challenge then is to sort through the possibilities and the associated technological and business challenges and identify opportunities that make sense for an MSO to pursue. This paper provides such a framework for pursuing opportunities with IoT. The framework that is proposed is focused on trying to build lasting value for the MSO by looking at the entire value chain for each opportunity and determining if it makes sense for the MSO to participate within that value chain

    See publication

Patents

  • Event Monitoring

    Issued 11314858

    A computing device may detect events such as a break-in, fire, flood, movement of people between different areas or zones within a defined area, cyberattacks, movement of devices away from the defined area, etc. If an event is detected, the computing device may take action to protect devices, data on the devices, and/or accounts accessible by the devices. The devices may encrypt, backup data, and/or delete data. The computing device may communicate with other computing devices about events that…

    A computing device may detect events such as a break-in, fire, flood, movement of people between different areas or zones within a defined area, cyberattacks, movement of devices away from the defined area, etc. If an event is detected, the computing device may take action to protect devices, data on the devices, and/or accounts accessible by the devices. The devices may encrypt, backup data, and/or delete data. The computing device may communicate with other computing devices about events that have been detected.

    Other inventors
    See patent

  • Methods And Systems For Delivering Content

    Issued US 9015274

    This disclosure relates to methods and systems for delivering content such as enhanced television applications over a network. In one exemplary method, a request for a content item is received. The request is added to a request queue. Each application requested is then identified and then the exemplary method waits for the occurrence of a predetermined event. Upon the occurrence of a predetermined event, the exemplary method provides for sending a single instance of each content item…

    This disclosure relates to methods and systems for delivering content such as enhanced television applications over a network. In one exemplary method, a request for a content item is received. The request is added to a request queue. Each application requested is then identified and then the exemplary method waits for the occurrence of a predetermined event. Upon the occurrence of a predetermined event, the exemplary method provides for sending a single instance of each content item, regardless of the number of requests for each such content item.

    See patent

  • Methods and systems for delivering content

    Issued US 9462064

    This disclosure relates to methods and systems for delivering content such as enhanced television applications over a network. In one exemplary method, a request for a content item is received. The request is added to a request queue. Each application requested is then identified and then the exemplary method waits for the occurrence of a predetermined event. Upon the occurrence of a predetermined event, the exemplary method provides for sending a single instance of each content item…

    This disclosure relates to methods and systems for delivering content such as enhanced television applications over a network. In one exemplary method, a request for a content item is received. The request is added to a request queue. Each application requested is then identified and then the exemplary method waits for the occurrence of a predetermined event. Upon the occurrence of a predetermined event, the exemplary method provides for sending a single instance of each content item, regardless of the number of requests for each such content item.

    See patent

  • Dynamic generation of a quantization matrix for compression of a digital object

    Issued US 8897586

    Some aspects of the disclosure relate to a compression technique that can permit determining dynamically a satisfactory quantization matrix based at least on properties intrinsic to a digital object being compressed and a predetermined compression quality criterion, wherein the quantization matrix is associated with a specific space-domain-to-frequency-domain transforms. In one aspect, the compression technique can permit creation of a compressed digital object that can satisfy a predetermined…

    Some aspects of the disclosure relate to a compression technique that can permit determining dynamically a satisfactory quantization matrix based at least on properties intrinsic to a digital object being compressed and a predetermined compression quality criterion, wherein the quantization matrix is associated with a specific space-domain-to-frequency-domain transforms. In one aspect, the compression technique can permit creation of a compressed digital object that can satisfy a predetermined a compression quality criterion.

    See patent

  • Dynamic generation of a quantization matrix for compression of a digital object

    Issued US 9479798

    ome aspects of the disclosure relate to a compression technique that can permit determining dynamically a satisfactory quantization matrix based at least on properties intrinsic to a digital object being compressed and a predetermined compression quality criterion, wherein the quantization matrix is associated with a specific space-domain-to-frequency-domain transforms. In one aspect, the compression technique can permit creation of a compressed digital object that can satisfy a predetermined a…

    ome aspects of the disclosure relate to a compression technique that can permit determining dynamically a satisfactory quantization matrix based at least on properties intrinsic to a digital object being compressed and a predetermined compression quality criterion, wherein the quantization matrix is associated with a specific space-domain-to-frequency-domain transforms. In one aspect, the compression technique can permit creation of a compressed digital object that can satisfy a predetermined a compression quality criterion.

    See patent

  • System and Method for identifying suitable time for pushing instantaneous donation requests and facilitating donation collection

    Filed Provisional Filing - 63270952

    System and method for pushing donation requests and facilitating donation collection are provided. In an embodiment, an electronic donation collection system allows a user to send donation request to one or more users. The system identifies an appropriate time for pushing the notification based on any or combination of parameters including but not limiting to the context of conversation proximity of the user, location of the user, and semantic analysis of a speaker’s voice. The system further…

    System and method for pushing donation requests and facilitating donation collection are provided. In an embodiment, an electronic donation collection system allows a user to send donation request to one or more users. The system identifies an appropriate time for pushing the notification based on any or combination of parameters including but not limiting to the context of conversation proximity of the user, location of the user, and semantic analysis of a speaker’s voice. The system further identifies one or more users based on user profiles with the purpose of the donation request for pushing donation request. The system may recommend a user to send the donation request to the one or more user at the appropriate time. In an embodiment, the system may automatically push donation requests, if configured a user to do so, to one or more users at the appropriate time. The system facilitates users to make an instant payment or agree to make a future donation using a user device.

Projects

  • Rumor Detection on Twitter

    -

    See project

  • Facial Keypoints Detection

    -

    Facial Key Points Detections using Convolutional Neural Nets

    See project

  • TLDR; Text Summarization using NLP

    -

    This paper explores the use of sequence to sequence networks [13] for abstractive text summarization tasks. The model overlays an attention mechanism and a beam search decoder on top of a vanilla Seq2Seq network model. Since Seq2Seq models are prone to grammatical errors and repetition, an n-gram based language model is then used to validate top beam search results and pick the best summary. The baseline model consists of a vanilla Seq2Seq model with attention and a greedy search based decoder.…

    This paper explores the use of sequence to sequence networks [13] for abstractive text summarization tasks. The model overlays an attention mechanism and a beam search decoder on top of a vanilla Seq2Seq network model. Since Seq2Seq models are prone to grammatical errors and repetition, an n-gram based language model is then used to validate top beam search results and pick the best summary. The baseline model consists of a vanilla Seq2Seq model with attention and a greedy search based decoder. The results show reasonable performance for the baseline and demonstrate the working of the language model as a noisy channel validation model. There were some issues overlaying a Beam Search decoder on top of the Bahdanau attention mechanism due to potential implementation issues. It is possible that with more training time and/or tweaks to the code, the beam search model performance complemented with the n-gram based language model will outperform the greedy search approach as expected. Overall however, this paper and project were immensely valuable in enhancing my understanding of Sequence to Sequence networks mechanism for text summarization and the Tensorflow development framework.

    See project

Languages

  • Tamil

    Limited working proficiency

  • Hindi

    Limited working proficiency

  • Telugu

    Limited working proficiency

  • Spanish

    Limited working proficiency

  • English

    Native or bilingual proficiency

Recommendations received

13 people have recommended Sanjay

Join now to view

More activity by Sanjay

View Sanjay’s full profile

  • See who you know in common
  • Get introduced
  • Contact Sanjay directly
Join to view full profile

Other similar profiles

Explore collaborative articles

We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.

Explore More

Add new skills with these courses

See all courses