Amit Gupta

Experienced Senior Software Engineer
An experienced software engineer with broad domain expertise spanning Operating Systems and Distributed Systems, Big Data systems, Machine Learning, Deep Learning and Web Development

Languages and Frameworks

Proficient: C, C++, Python
Frameworks: Tensorflow, Pytorch, OpenMP, MPI, Pthreads, Hadoop, HIVE, Spark, Django
Prior Experience: Java, SQL, R, Linux Kernel Programming

Experience

2013-Now

Research Software Engineer
Texas Advanced Computing Center
Austin, Texas

Project: Scalable Object Detection and Analysis
- Implemented end-to-end computer vision pipeline to detect and track pedestrians in traffic camera videos
- Implemented standalone C++ applications for pedestrian detection and tracking
- Pedestrian Detection and Tracking in Traffic Camera Videos.
- Used the darknet framework to implement object detection and tracking pipeline
- Implemented code to provide statistical summaries of pedestrian crossing event counts
- Data from 4 traffic junctions processed and analysed through the developed pipeline
- Analysis data used by City of Austin in their infrastructure/urban planning efforts
- Technologies and Tools Used
  - C, C++, Darknet (Neural Network Framework),
    bash, Python
Project: Domain Informational Vocabulary Extraction (DIVE)
- Designed and implemented DIVE, a tool for entity extraction from textual content of technical articles using NLP techniques
- Implemented web interface for entity curation by expert users.
- Implemented a web service for integration with publication pipeline of 2 journals.
- Deployed DIVE as a live service used by the publication pipeline of 2 prominent journals from Oxford University Press (The Plant Cell and Plant Physiology)
- Implemented curation interface to help authors curate data from their own articles
- Implemented the use of association analysis to detect relationships between entities based on their co-occurance patterns in text.
- Created a gold standard dataset for training Neural Networks using author curation inputs submitted to DIVE.
- Explored Deep Learning based approach using NeuroNER library (that implements LSTM networks) to improve accuracy of entity detection. Showed a 4x improvement in performance in comparison with other methods.
- DIVE is under ongoing development and we are exploring reducing the training time of explored the explored Deep Learning methods to facilitate its inclusion into the tool. We are also adapting DIVE to mining corpora in other domains like Aerospace Engineering.
- Technologies and Tools Used -Java, DJango/Python, Python, Tensorflow, NeuroNER (Neural Networks for entity extraction)
Project: AstriaGraph Project
- Implemented use of graph databases (Neo4j) for tracking man made space objects
- Contributed to development of new data model
- Implemented data ingest pipeline to ingest data from 7 data sources
- Implemented cesium based visualization of space objects with orbital parameters
- Implemented several front-end, proof of concept use cases
  - UNO Compliance: Tracking owning entity (i.e space agency/country, company) compliance with UNO space object registry guidelines
  - Visualization of schema design of the data model
- Technologies and Tools Used
  - Neo4j, Javascript, Cesium and javascript visualization libraries
Project: Scaling Traffic Simulation software VISTA on HPC machines
- Ported VISTA (a popular city scale, Dyanmic Traffic Assignment based Traffic Simulation and Analysis tool used by transportation researchers) to Intel architecture for upto 50% improvement in runtime
- Parallelized cost update function in VISTA, resulting in improved solution search
- Redesigned and Implemented parallelization changes and resolved memory bottlenecks in VISTA for scaling on HPC infrastructure
- Designed and implemented a priority queue based heuristic to optimize a time-dependent shortest path algorithm in VISTA to significantly improve worst case outliers
- Improved overall VISTA application performance by 6x-12x.
- Ported VISTAs cost update function to Intel MIC platform
- Technologies and Tools Used
  - C++, ,Pthreads, MPI

Education

2010 - 2013: Master of Science (MS), Computer Science (w/ Thesis)
GPA 3.81
Thesis "Motivating Elastic Operating Systems in the Cloud"
University of Colorado Boulder
2005: Post Graduate Diploma, Embedded Systems Design
Advanced Computing Training School
Center for Development of Advanced Computing (C-DAC), Pune, India
1999 - 2003: Bachelors of Engineering, Electrical & Electronics Engineering
First Class with Distinction
Visveswaraiah Technological University, India

Publications

: Google Scholar Profile

Work Authorization

: US Permanent Resident
(No work visa sponsorship required)

gupta.v.amit@live.com •+1 512 537 6418•
7714 Louis Pasteur Dr, 1318, San Antonio, TX 78229
agupta.net