This is an index of publicly-available projects I’ve worked on.
Open Source Contributions
Started November 2019
Elastiknn is an Elasticsearch plugin for exact and approximate nearest neighbor search on numerical vectors. It allows Elasticsearch developers to combine traditional queries (e.g. documents containing “some product”) with vector search queries (e.g. documents with images similar to some image).
Talks and Presentations
Using Elastiknn for Exact and Approximate Nearest Neighbor Search
December 2020 – Elasticsearch Online Meetup
Meetup presentation covering the Elastiknn plugin.
Keras Library for Neural Networks and Deep Learning
June 2017 – Knoxville Data Science Meetup
Meetup presentation covering various aspects and use-cases for the Keras library.
Exploring Serverless Architectures with AWS Lambda and Node.js
July 2016 – Codestock Conference
Conference presentation introducing the AWS Lambda service with a simple web application implemented on Lambda.
React, Flux, and Realtime RSVPs
Meetup presentation where I live-coded a web-application that uses React and D3.js to display statistics about RSVPs from the Meetup.com streaming API.
Kaggle KKBox Music Recommendation Competition
October - December 2017 – Competition
Used collaborative filtering, vector-space models, and gradient-boosting classifiers to train a music recommender system. Submission ranked 222nd (top 21%).
Kaggle Understanding the Planet from Space Image Classification Competition
May - July 2017 – Competition
Trained convolutional networks to do multi-class image classification on satellite images. Submission ranked 103rd (top 11%).
Neurofinder Calcium Imaging Segmentation Challenge
April - July 2017 – Competition
Trained a fully-convoluational neural network to segment neurons from images of a mouse brain taken using a calcium-imaging technique. Placed third in the Neurofinder competition. Published results at the 2017 Deep Learning in Medical Image Analysis Workshop.
ISBI 2012 Electron Microscopy Segmentation Challenge
March 2017 – Competition
Trained a fully-convolutional neurla network to segment electron microscopy images. Placed top 50 on leaderboard (search “ORNL_Klibisz”).
Reservoir Computing Summary Paper
Term-paper for Biologically-Inspired Computing course at the University of Tennessee. Covers Reservoir Computing, Liquid State Machines, and Echo State Machines.
Meetup RSVP Visualization
Real-time visualization of streaming data from the Meetup.com open events RSVP API. Visualizes events as a map with markers and bar charts showing the most frequent countries, states (USA), and names.
Dynamic Adaptive Neural Network Array Visualization
The Dynamic Adapative Neural Network Array (DANNA) is a neuromorphic computing model created by the Neuromorphic Computing research group. I built DANNA-View, a web-app for visualizing the DANNA networks, as an independent study course with Dr. James Plank. The tool produces interactive visualizations and heat maps for network activity, which are used for understanding and debugging the networks. The app uses a node.js server to transform network definition files to a JSON format for rendering, and React.js and D3.js in the browser to render the network and provide controls for interacting with it.
Animated Finite State Machine Converter
An application that lets the user interactively build a Nondeterministic Finite Automaton and visually convert it to an equivalent Deterministic Finite Automaton. Built as an honors project for the spring 2015 COSC312 course at the University of Tennessee. Used D3.js and Angular.js to create the visualization and controls.