Close

Ishank Sharma

AI Solution Architect

Download Resume

About Me

I graduated from Rutgers University, New Brunswick majoring in Electrical and Computer Engineering. I have a unique interdisciplinary research exposure to AI systems. My projects range from designing machine learning based risk assessment models, vision systems for autonomous vehicles, all the way to studying emergent phenomenon in complex systems.
I explore new drum grooves and like to read in my downtime.

Experience

CYR3CON Inc., AZ

AI Solution Architect

  • Led development of NLP infrastructure and products for predicting/recognizing exploits, vulnerabilities, attack patterns.
  • Lead improving data and prediction pipeline for 5X faster running performance and increased request serving capacity by 60%.
  • Designed and developed systems for automated CICD and performance evaluation, presenting to directors and C-levels regularly. Managing 4 other data scientists, analysts and reporting to the CTO.

Verizon, NJ

Software Engineer

  • Developed rapid prototypes to provide proof-of-concept for validating new approaches for 5G network planning in United States.
  • Built deep learning models, analytics pipelines and sofware applications to generate/analyze 5G coverage maps using PostGIS, SQL, AWS EC2, Javascript, Android applications

State Street Corp, TX

Cloud Infrastructure Engineer

  • Developed CI/CD pipeline to provide Cloud Infrastructure as a Service. Worked with underlay team to develop automated testing environment for cloud service controller components.
  • Responsible for maintaining/monitoring deployed cloud compute and storage instances.

State Street Corp, NY

Data Science Co-op, Verus

  • Combined natural language processing, deep neural networks and financial risk models to find connections between global news, risk events and multi asset portfolios streams for investment decision support. I worked on tasks ranging from topic modeling, unsupervised contextual clustering, knowledge based entity linking, ARIMA-LSTM financial time series analysis.
  • Developed RESTful web services/interactive dashboards for visualization and analysis of deployed machine learning models

Project Link: State Street Verus.

Rutgers Univeristy, New Brunswick

Graduate Teaching Assistant

  • Worked as a Teaching Assistant for Cloud Computing Fall 2017 (taught by Professor Maryam Mehri Dehnavi).
  • Responsible for developing course content, guiding and supervising student teams.

Intutent Inc., Delhi

Software Engineer

  • Created ETL infrastructure for BigData based services with Python, Java, SQL, MongoDB, Redis, Django and SparkML.
  • Developed deep neural network based QA chatbots to automate job interview sessions.

Indraprastha Institute of Information Technology, Delhi

Research Associate

  • Responsible for developing natural sounding speech synthesis system using hindi transcribed texts. I used RNN for duration/acoustic modeling and WORLD vocoder for waveform synthesis.
  • Worked with vision team to design pedestrian detection system for an autonomous driving vehicle using convolutional neural networks

Project Link: Swarath & I am Kalam.

Education

Rutgers Univeristy, New Brunswick

Sept 2017 - May 2019

MS in Electrical & Computer Engineering

GPA: 3.2/4.0

MS Thesis: Using News and Supply Chain Data to Predict Stock Price Movements.
Coursework: Introduction to Computational Robotics, Computer Vision, Statistical Learning, Convex Optimization for Engineering Applications, Data Structures and Algorithms, Machine Learning, Mobile App Engineering with Android, Programming Finance with C++, Cloud Computing

Activities:

USICT, Guru Gobind Singh Indraprastha University, Delhi

Aug 2012 - July 2016

BTech. in Electronics and Communication Engineering

GPA: 3.59/4.0

BTech Final Project: On the role of evangelism in consensus formation: a simulation approach.
Coursework: Transmission Lines, Quantum Physics, Engineering Mathematics, Numerical Methods, Control Engineering, FPGA Design and Logic, DBMS with SQL and NoSQL, Object Oriented Programming with C++, CMOS Circuit Design, Mobile Communication, Computer Architecture & Operating System, Network Design, Software Engineering, Algorithms

Activities: Senior Student Advisor, Bio-inspired Computation Lab, USICT

Publications:

Projects

Motu-Petu: A Food Recognition App

  • The android application allows users to take picture of the targeted dish and then recognizes the dish from our database of hundreds of different food items which includes american, chinese, korean and japanese cuisines. The user can also share the clicked photo on his facebook account.
  • The app utilizes a mobile efficient Convolutional Neural Netowork for fast inference. We trained the network on 139400 food dish images and 357 food categories and obtained Validation Top-1 accuracy of 44.1% and Test Top-1 accuracy of 43.7%.
View Project

Interactive user marked segmentation with GraphCut

  • Developed an interactive GUI based image segmentation tool (in MATLAB and Python). The proposed algorithm grows foreground and background regions from user marked input seeds.
  • Applied Markov Random Field based image energy minimization and support vector machines for classification. Performance evaluation on binary segmentation tasks using BSD300 dataset.
View Project

Portfolio optimization with linear and fixed transaction costs

  • Developed optimal portfolio selection algorithm maximizing the returns over a period of time with linear and fixed transaction costs, constraints on exposure to risk, short-sell constraints etc
  • Pertubation analyis of alpha on 10 year stock data of 100 tickers.

Unsupervised Segmentation of Food Images

  • Developed Convolutional Neural Networks for unsupervised instance segmentation of food items in a plate.
  • Implemented Region proposal Networks, Pyramid feature parsing and W-Net architecture for improved localization of segmentation mask.
View Project

Distributed computing for Machine learning

  • Developed machine learning models for network intrusion detection and object image classification on a distributed system using Hadoop, Spark, SQL, Kafka, Scala, Java, MLlib.
  • Presented tutorials in ECE579 course. Topics included applications of MapReduce and Hive/SQL datawarehousing

Representing contextual relations in ancient Sanskrit texts

  • Proposed and implemented an algorithm to parse/represent vector embedding of Devanagari words in ancient Sanskrit literature- Ramayana, Vedas, etc. Obtained word vectors for 39940 most frequent words
  • Developed stacked Bidirectional LSTM network with word vectors as embedding layer weights for dense representation of large input vocabulary.
  • Proposed LSTM architecture with residual skip connections to learn and predict complex relationships in Sanskrit Devanagari texts.
View Project

Skills

Get in Touch