Manasa Kumar

3rd

Previous positions

  • Teaching Assistant at University of California San Diego
  • Research Intern at Indian Institute of Technology, Madras

Education

University of California San Diego, Master’s degree, Computer Science

477

Background

Experience

  • Teaching Assistant

    University of California San Diego

    September 2019 – Present(1 month)La Jolla

    TA for CSE 21: Mathematics for Algorithms and Systems Analysis.

  • Software Engineer Intern

    Amazon

    June 2019 – Present(4 months)Greater Seattle Area

    + SDE Intern in the Co-branded Credit Cards Team of Consumer Payments. + Introduced ability for customers to add authorized users with Amazon co-branded credit cards purchase. + Modified full-stack in Spring MVC and Java to include functionality for supporting the new use cases.

  • Teaching Assistant

    University of California San Diego

    January 2019 – June 2019(5 months)La Jolla

    TA for CSE 20: Introduction to Discrete Mathematics. TA for CSE 21: Mathematics for Algorithms and Systems Analysis.

  • Research Intern

    Indian Institute of Technology, Madras

    January 2018 – May 2018(4 months)Chennai Area, India

    + Code-Mixed Question Answering under the supervision of Dr Mitesh M. Khapra. + Collected 50K code-mixed Hindi-English QA data points through Amazon Mechanical Turk. + Implemented a model that learns character and contextualized word-level representations using Tensorflow. + Achieved an accuracy of 93% for word level language identification in the code-mixed text. + Implemented the QA models – Bidaf and R-Net in Tensorflow. + Compared the results of QA models on the code-mixed and English data.

  • Summer Research Intern

    Indian Institute of Technology, Hyderabad

    May 2016 – July 2016(2 months)Hyderabad Area, India

    + Worked with Dr Vineeth N. Balasubramanian on applying Machine Learning techniques to agricultural datasets in collaboration with Dr Wei Guo at The University of Tokyo. + Project titled ‘To Go Deep Or Not To : Detection of flowering panicles in images of paddy-rice’. + Obtained 92% accuracy and 0.54 correlation using SIFT descriptors with an SVM classifier in Keras. + Increased the accuracy to 99% and correlation to 0.74 using a deep learning feature extractor. + Experimented with an entropy based Active Learning approach to increase efficiency.

Education

Languages

  • English

  • Tamil

  • Sanskrit

Projects

  • Reconstruction of right half of MNIST image given left | Tensorflow

    + Implemented self and cross reconstruction with Correlational Neural Nets. + Experimented with extending this model to cross-modal reconstructions.

    Team Members (1):
    • Manasa Kumar
  • Video VQA with TGIF-QA dataset | Caffe

    + Obtained C3D and ResNet feature vectors for the TGIF-QA dataset for the ST-VQA model.

    Team Members (1):
    • Manasa Kumar
  • Sleep Apnea detection using RNNs based on ECG signals | Keras

    + Trained an RNN to classify sleep apnea from ECG signals obtained using a pulse oximeter.

    Team Members (1):
    • Manasa Kumar
  • Movie recommendation based on common genre relations | Python

    + Formed co-occurence graphs on the MovieLens dataset to model genre relations between movies. + Used random walk algorithm on graphs to recommend movies to users with F1 score of 0.25.

    Team Members (1):
    • Manasa Kumar
  • Melanoma classification from images of skin | Keras

    + Experimented with different deep learning feature extractors for image representation. + Obtained 74% accuracy using the Inception V3 features with an SVM classifier.

    Team Members (1):
    • Manasa Kumar

Skills & Expertise

  • CSS3
  • Keras
  • JavaScript
  • Bootstrap
  • Python
  • jQuery
  • TensorFlow
  • PHP
  • AngularJS
  • C
  • Django
  • OpenCV
  • MySQL
  • Caffe
  • C++
  • Scikit-Learn
  • Pandas