Purvak Lapsiya
lapsiya at usc dot edu

I am a second year computer science MS student at University of Southern California, researching at USC ICT. I also collaborate with the Associate CTO of NASA Jet propulsion laboratory Dr. Chris Mattmann at Information Retrieval & Data Science Group. I earned my Bachelors in Computer Engineering in 2018 from University of Mumbai,India.

I'm currently working at an early stage startup frenzy.ai. I have had the fortune to intern / conduct research at SAP, Palo Alto (Summer 2019) mentored by Dr. Amit Nagar and; Indian Space Research Organization, (Spring 2018), on a diverse set of topics - ranging from vision & language to UAVs.

Email  /  CV  /  LinkedIn  /  Github  /  Twitter

profile photo
Research

I am interested in the broad areas of Machine Learning, Computer Vision & Language. My primary research interest is in "understanding" models in context of fairness and to make models "understand" environment and come up with fair decision. My undergraduate thesis Personal Assistant Drones with Semi-Autonomous Flight Capabilities extended the concept of developing software suite for UAVs which has swarm capabilities. I'm also intreseted in fast semantic segmentation and semi-supervised learning.

News
  • [Jan 2020] I will be interning at frenzy.ai in Los Angeles
  • [Dec 2019] We won the Best Paper Award at I/ITSEC'19, the world's largest modeling,simulation and training conference!
  • [Dec 2019] Presented our work on "Fully Automated Photogrammetric Data Segmentation and Object Information Extraction Approach for Creating Simulation Terrain " at, I/ITSEC 2019.
  • [May 2019] I will be interning at SAP at Palo Alto in summer 2019.
  • [Jul 2018] I will be joining University of Southern California as a Masters of Computer Science student in Fall 2018.
  • [Mar 2018] Runner's up Smart India Hackathon'18 under department of space!.
Research/Work Experience

3DSP Software Engineer Intern, Frenzy.ai
Building Infrastructure for Machine learning production.
3DSP Data Scientist Intern, SAP
Designed an intelligent service detect custom tables in PDF documents and extract information for clients who processes 1,000,000 invoices.
Researched on improving fairness of job description by removing unconscious gender biased words.
3DSP Graduate Research Assistant, USC Institute for Creative Technologies
Researched on learning dense volumetric segmentation from sparsely annotated point clouds in a semi-supervised manner for better simulation and modeling of 3D terrain.
Implemented distributed training and re-training procedures for a neural network to handle 18 million data points per point cloud.
[Media Coverage]
3DSP Directed Research, IRDS
Developed a fine-grained image similarity using deep ranking to improve image retrieval of similar images for memex.
[Github]
3DSP Software Engineering Intern, Indian Space Research Organization
Trained and built CNN based classifiers using tensorflow for Garbage, Pothole and Road detection with an accuracy of 92%.
Assisted with the development of node.js application dashboard and specifically worked in facilitating visualizations of swarm status and node heartbeats.
Optimised traditional master-apprentice fashion by introducing relay points for timed data collection, minimizing data packet drops close to 0.
3DSP Research intern, K.J. Somaiya College of Engineering
Created an online training system which gauges the understanding of learner’s emotions in the context of learning.

We will not only use the machines for their intelligence, we will also collaborate with them in ways that we cannot even imagine. ~ Fei-Fei Li (Design and CSS courtesy: Jon Barron, Amlaan Bhoi and Purva Tendulkar)