About Me, and my interests

Who Am I?

Hi I'm Ananya I am a Masters student at University of California, Irvine. I have worked on deep neural networks in field of biomedical imaging before,high energy physics and gender bias in language. Areas I'm broadly interested in:

Machine learning

Deep learning

NLP

Image Processing

Education

Education

2016-2018, GPA: 4.0/4.0

Cups of coffee
Projects
Companies
GPA
Experience

Work Experience

Data Science Intern - Microsoft Summer 2018

Azure Customer Growth and Analytics
  • Achieved 7x improvement over previous version of Azure customer churn prediction model
  • Increased granularity by expanding scope to multiple services, finer customer level, and increased time frame
  • Deployed classical ML algorithms (xgboost etc.), extended to LSTM-based network to include time series data (python)
  • Worked with vendor team to ensure data quality and correctness, focused efforts on feature design and selection

Data Science Intern - Microsoft Summer 2017

Azure Customer Growth and Analytics
  • Built first model for Azure service recommender system
  • Posed the problem as combination of multi-class multi-label classification, and user-cluster-based classification
  • Model deployed in production; showcased at MLADS (Machine Learning and Data Sciences) conference at Microsoft

Graduate Student Researcher - UC Irvine Jan'17 - Jun'17

  • Extracted author-gender information automatically from political science reading lists and syllabi
  • Deployed CRFs to find dierent parts of citation, achieved accuracy ~93% for dierent parts

Data Science Intern - Qure.ai May'16-Jul'16

  • Achieved 95% accuracy for automatic segmentation of 3D knee MRI images into bones and cartilages
  • Designed a U-net convolutional neural network combined with dilated convolutional networks in torch (lua)
  • Time taken per image 8s (15 min required by a practitioner); reduced to 1% of human time

Software Engineering Intern - CERN Oct'15-Mar'16

  • Analysed cache behavior using Valgrind and VTune, changed data layout in memory for 1.3 times quicker access
  • Achieved speedup of 2.5 by deploying instruction level parallelism(C++) using SIMD registers in a bilinear interpolation routine to calculate magnetic field in high energy physics reactions in ATLAS detector

Research Assistant - IIT Bombay Jul'15-Sep'15

  • Empirically improved bound on transience class of n*n grid for Abelia sandpiles model from O(n7) to O(n3)
  • Implemented and analysed Integer Linear Programming algorithms (python) to find best augmenting path in Ford Fulkerson algorithm for max flow/min cost in a graph network; lesser iterations than Edmonds Karp
Publications

Publications and Posters

Posters

Ananya, et al., Sameer Singh. Automated Detection of Gendered Text (Poster at WiNLP, ACL 2017 and WiML, NIPS 2017)
Pipeline for automatic detection and quantification of gendered text for dierent domains; performance comparable to humans-evaluated using MTurk



Publications

Google Scholar profile