As a curious scientist and engineer, my interests are twofold. For one thing, I love studying and modelling natural systems. Such models expose the underlying structure inherent to natural systems, that we fail to observe otherwise. On the other hand, I also deeply enjoy modelling, designing and implementing computational systems inspired by natural systems. riyansh.karani.011235@gmail.com +1 (872)-333-9826

var riyansh_karani = function() {

// Hello, you have found my website.

this.name = "Riyansh Ketan Karani";

this.location = {

// University of Illinois at Chicago

// Chicago, Illinois - 60607

var (latitude, longitude) = (41.871251, -87.652983);

}

this.is_happy = function() {

if (this.is_coding()) {

// I am always happy when I am coding :)

return true;

} else if (this.is_learning()) {

// Above everything else, learning is my passion

return true;

} else if (this.is_travelling()) {

// Travelling helps me explore the world, put my

// life in perspective (and also keeps me away

// from my Laptop's screen for a while).

return true;

} return false;

}

// Life Motto

if (!this.is_happy()) {

this.be_curious();

this.be_awesome();

}

// ...

}

I am an Electronics Engineer, a Software Developer and Computer Scientist. I live in the windy City of Chicago. It's always cold here! I view my life as a Markov Chain, wherein the edges represent the choices that I have made. In that case, determining the so called goodness / badness of a choice is just a simple inference problem, based on the outcomes (nodes) in this chain. So, when I say that I wish to pursue a career as a Researcher in the field of Computer Science, it is only a decision under uncertainty that maximises the number of the experiences in life that give me joy and satisfaction. I am excited about research in fundamental areas of Artificial Intelligence. I have come to like probabilistic models in general and graphical models in particular, not only because they provide a powerful way to model complex stochastic systems but also from the algorithmic point of view.

Achievements

MIT Energy Hackathon 2017

Purpose Energy is a company that uses their Tribrid-Bioreactor biogas plant to reuse industrial waste and generate methane rich biogas that can in turn be used to power a Cogen engine that generates electric power. We worked over a period of 36 hours to come up with modifications to their existing Tribrid-Bioreactor design and propose a control system that can automate and optimize the efficiency of their bioreactor to run at a 100 percent efficiency irrespective of the type, concentration and amount of substrate (waste) that was dumped into the bioreactor.

My contribution to the project was to come up with a 2 staged neural network architecture that first models the biogas plant as a state machine, by training a Recurrent Neural Network to emulate the reactor. This Recurrent Neural Network (RNN) is then used to emulate the biogas plant and a simple Multilayered Neural Newtork is trained to control the output of the RNN (the state variables) to drive the system from some random state to the desired state, by minimizing a specifically designed error function.

Thanks to Dave Amirbekyan with his terrific Chemical Engineering skills and expertise.


Projects

notebook.js

A Javascript framework for building websites that resemble a notebook. In fact, this entire webpage is written in pure HTML and all the styling, animations, responsiveness, distributing information in separate columns etc. is all being automatically handled by notebook.js.

The basic aim here was to build a javascript framework that could convert any existing HTML webpage into a notebook type page, and make it extremely easy for one to add sidenotes to the main content. (In fact, this webpage is written as a static html page, and all the placement of content, sidenotes and animations have been handled by notebook.js).

Real time Handwritten Character Recognition using Image processing and Neural Networks implemented on an FPGA device

Implemented all image processing from scratch in C, and implemented C-python binding to access the C Library from Python. Designed and implemented a Neural Network classifier in Verilog for a Spartan 6 FPGA device.

This was my final year project where I and my three other teammates worked for an year to build the entire system from scratch.

Recurrent Neural Network based pen for Handwritten character recognition

Designed and developed an Inertial measurement unit (IMU) based embedded system to obtain orientation data a pen with the IMU attached to it. Implemented a simple Recurrent Neural Network and trained it to learn and detect sequences in IMU data as representing one of 36 possible characters (A-Z and 0-9).

This was my 6th semester project. Implemented the entire embedded system from scratch, and used probabilistic models in real life applications for the first time.

4 - Axis SCARA Robot

Worked on building the electronics, hardware and software for an interdisciplinary project involving building a 4-axis SCARA robot.

This was the longest and most irritating (at times) project I have ever worked on. But, when things worked, it felt like heaven :)

VLSI Chip Design

Completed schematic entry, layout and corner simulations (with pad frame) for the following circuits using AMI 0.5 micron technology :

Did this as a part of a 3 month summer training program.


Publications

Riyansh Karani, Akash Rana, Dhruv Reshamwala, Kishore Saldanha (2016) “A floating point division unit based on Taylor-series expansion algorithm and Iterative Logarithmic Multiplier”. In Proceedings of the Second International Conference on Computer Science, Information Technology (CSITEC -2016).

To reduce the size of the PDF document, the editors of the journal took screenshots of parts of the paper and added the images in the published copy, because of which parts of the published version are blurry (including all the images). The original version is much more crisp, if you are interested.


Work

Full stack developer, Scandid.in

Scandid is an e-commerce company located in Pune, India, where I worked as a part of the backend team on the database management system and web crawler system. Additionally, for a brief period, worked with the fronted team on their android platform, and the Google Chrome Extension platform.


Research Intern, Computer Science and Engineering Department, Indian Institute of Technology Bombay

Working as a research assistant under the guidance of Professor Ganesh Ramakrishnan at the Computer science and Engineering department at Indian Institute of Technology Bombay.

  • Worked on building a desktop and android apps for Lokavidya, a social media platform for rural tribal people in the state of Maharashtra. During this project, worked on a video classification problem for converting hierarchically modelled data to Pachinko allocation model.
  • Working as an android developer for Lokacart, an e-commerce platform which aims to restore farmers as the primary vendors of farm products, by providing a direct line of communication between the buyers and the farmers.
  • Working on a research project that aims to use Deep learning and signal processing to build audio-visual saliency models for static video summarisation.