Publications

Block Propagation Applied to Nakamoto Network Link
  • Compares the performance of Compact blocks and xThin blocks in the Dash network
Velocity Link
  • Scalability Improvements in Block Propagation Through Rateless Erasure Coding
  • Introduces a new protocol to relay block data by aggregating partial information from multiple peers.
  • Introduces potential to relay block information over UDP networks.
Digital Fountain for Multi-Node Aggregation of Data in Blockchains Link
  • Introduces a technique to relay block data by aggregating partial information from multiple peers.
  • Introduces potential to relay block information over UDP networks.

Experience

Blockchain Engineer

01/2019 - 09/2019
IBM Blockchain - Worldwire
  • Infrastructure
    • Designing and implementing and deploying a microservice architecture for Worldwire using kubernetes, istio service mesh
    • Automated complete deployment for all microservices and incorporated all states in a service deployed in google app engine
    • Designed and implemented auth service for the client side authentication and authorization in golang.

Graduate Service Assistant

08/2018 - 12/2018
Arizona State University
  • Thesis Digital Fountain for Multi-node Aggregation of Data in Blockchains
    • Defended thesis, introducing and evaluating a new block propagation that utilizes rateless erasure codes in order to retrieve block data.
    • The study finds that this on dash blockchain the network will maintain consensus and remain profitable for miners block of sizes upto 650MB, and will make xthin/graphene/compact more capable for block propagation by aggregating block data from multiple nodes.

Software Developer Intern

05/2018 - 08/2018
Dash Core Group
  • Graphene Block propagation protocol GitHub
    • Graphene utilizes bloom filters, inverted bloom look up tables and transaction order information to optimize block propagation
    • Using the Bitcoin Unlimited implementation to add bloom filters and inverse bloom look table to the dash implementation.
    • Adding Expedited blocks implementation to dash.
    • Implementing RPC tests on dash core using ctypes and python binding for c++ for integration testing
    • boost c++ tests for writing unit tests

Research Assistant

08/2017 - 05/2018
Blockchain Research Lab, Arizona State University
  • Developed Dash-Simulator using Network Simulator 3 to help scale Dash to the next block propagation technique and block size with profitable orphan rates – dash-simulator
  • Added Compact and xthin block propagation techniques to the dash-simulator.
  • Tested traditional, Compact and xthin for 750 (for better confidence interval) blocks each and published results with orphan rates.
  • Utilized above knowledge about propagation to design a new protocol that utilizes fountain codes to propagate blocks

Co-founder, Director Research and Development

08/2017 - present
Blockchain Innovation Society
  • The non-profit aims at educating public on blockchains and forming an open platform for blockchain resources.
  • Presentaions, lectures and paper discussions with students at ASU in order to help better understand blockchains

Developer Business Intelligence

07/2013 - 07/2016
Target Corporation India
  • Real Estate Database
    • Developed a full stack application using .Net, angularJs (1.3) and SQL server
    • Date warehousing, design and ETL for the complete data from microsoft excel and Teradata to Microsoft SQL Server
  • Block Group Assignment
    • ETL, Sqoop data from Teradata to Hive
    • Implementing R-trees to index data for various locations
    • Implemented Ray casting to understand location of users in order to understand the shopping pattern
    • Utilized python, pandas dataframes, pyspark for all implementation and automation was done using shell scripts

Projects

In-memory Spatial Database Indexes
  • Developed PR-Quadtrees to index the space for the in-memory database.
  • Placed the pointers of a child quadnode contiguously in a vector collection instead of having one pointer to node like the classic implementation. This made the QuadNode have a smaller memory footprint in the cache and improved data locality.
  • Improving data locality resulted in descendants present in the cache increasing the cache hit-cache miss ratio.
  • For MXCIF Quadtree, the leaf capacity was made configurable, in order to reduce the fan out on insertion of new data.
  • The approach overall boosted the space usage by approx. 40.
  • Technologies: C++.
KDD Cup 2016
  • Predicted the relevance of an academic institution based on Microsoft Academic Graph data.
  • Generated statistical features by parsing and processing 120 GB of textual data.
  • Used Gradient Boosted Decision Trees, Ranked SVM to make predictions.
  • Technologies: Python, Google Big Query, Bash.
Lamport Clocks GitHub

Skills & Proficiency

C++, C

Python

JavaScript

Git, SQL

Docker, MongoDB