WORK

Machine Learning Intern at FJWing Health Technologies LLC

The overall goal of this project is to enable a pregnant woman to be able to know if her fetus is healthy within the comfort of her home and without the need of a doctor or a clinician.

In order to do that, we created an Android Application that receives the fetal heart rate and classifies the data as good, bad, or average. In order to get the application to tell a pregnant woman if her fetus is health, unhealthy or moderate, we are using machine learning algorithms classify the data. Give that the raw FHR data is obscure in the nature of what it is describing, it is necessary to extract important features from this data in order to classify. Two methods of feature extraction were implemented in this project. The first method, element-wise feature extraction, extracted important characteristics of the FHR from each element individually. The second method, real-valued feature extraction, created a feature set that describes the entire feature as one vector. With both of these feature sets, we implemented two machine learning models which were the Gaussian Mixture Model (GMM) and a Support Vector Machine (SVM) on MATLAB.

Of the two machine learning models, the accuracies we received are as follows:
SVM on Element Wise Feature Sets - 93%
SVM on Real Feature Sets - 76%
GMM on Element Wise Feature Sets - 50%
Once we got these results, we knew that we had to proceed with SVM. To do this, we got SVM running in an Android environment and wrote the feature extration code in C++.

Below is an image of how the whole system works.

Now, the completed product is being launched with the companies Android Application in China in early July, 2016.

Intern at Way.com

Last summer at Way.com, I got to help with developing and testing an 'Internet Of Things' device that the company was building for its users. The application that the company is currently going to use it for is parking garages and solving a problem that a lot of customers face. Currently when going into a parking garage, the user has to open his or her window, grab a ticket and then the gate proceeds to open. It is usually very inconvenient because it takes quite a bit of time to get the ticket and usually invovles a lot of uncomfortable stretching from the users. Also, after the user is done using the garage, he or she has to pay for the ticket using a machine and then only can they leave the garage.

The device that was developed and tested is attached into the parking garage gate. It allows for the user to connect to the garage using bluetooth from his or her phone and the app that the users will have to download. Once the user is connected, he or she can open up the garage gate and park the car. Once the user is done using the garage, he or she continues to leave the garage using the phone app. Once the user leaves the garage, the backend of the app will compute the amount of time that the user was in the garage and charges the user's bank account for the time that the user was in the garage.

Unfortuantely at this time, I am unable to give any more details about the device since it is not out in the market yet and is still being tested by others. Feel free to email me if you want more specifics and I can hopefully answer your questions.

Intern at BGR

During the summer of 2014, between the months of July and September, I interned at BGR energy systems in Chennai, India. While at BGR, I had the opportunity to work with the electrical projects department of the company. I was assigned my own project which was to design a smart grid network within the state of Tamil Nadu. As the literacy rate has increased and more of the population have college degrees, the average consumption per capita has severly increased. To deal with this increased demand for electricity, more power plants are being built across the state. The smart grid will allow for the state to get one step closer to having unmanned power plants where everything can be monitored and controlled from a central location.

The smart grid was designed using Optical Ground(OPGW) Wires and they would replace the ground wires on power lines. OPGW wires act as a ground wire as well as allow for information to be transmitted at the speed of light. The best way to go about designing the smart grid was to use a ring network as shown in the image below.

The ring netowrk allows for any power plant that is having any issue to be temporarily shut down till the problem is solved while redirecting electricity through the other sub-stations. The list power plants that are in the main ring and sub-stations in the sub-ring can be found here. The connections of the network from west to east as well as other single lines where the OPGW cable will be installed is included in the excel sheet as well. Unfortunately I do not have a picture of the ring with all the power plants and sub-stations with me.

Impact Team Memer at B2U (Currently Giventure)

Bystanders 2 Upstanders (B2U) was a startup that I was part between the months of November 2013 to January 2015. The start-up was founded by UCSD students and funded by Microsoft and I joined about 3 months after the start-up was initiated. The point of the start-up was to make a website as well as an app where people could find volunteering opportunites near them and if none were available, we made some challenges that people could do by themselves. Each task/opportunity had a difficulty rating and based on how difficult it was, the users got a certain number of points. Each point was worth a certain amount of money and the user could choose what organization they would like to donate their points (money) to. We thought we could attract a lot of users because one good action that the user does can have double the impact with the point system.

My role in the start-up was to see the best way to make an impact on the community. We had to come up with 10 challenges that people could do if they did not want to volunteer and with each challenge, I had to give it a difficulty rating and how many points it would be worth. In the year that I was with the start-up, I came up with around 200 challenges for the users and worked with Microsoft Azure to upload the challenges. I was also responsible for finding volunteering opportunities that people could do all around the world and made a list of those.

PROJECTS

Project Lead Engineers Without Borders UCSD

I have been the project lead of the Rwanda Project for more than a year now. The Rwanda Project is an EWB approved project that UCSD is doing in Bungwe, Rwanda. Bungwe is a rural farming town located in Eastern Rwanda and the main problem that the people of Bungwe are facing is a lack of drinking water. They have a reservoir which is about 500 meters below the village but getting the water from the reservoir to the village is a problem. There are a few people in the community who make a living out of taking jerry cans down to the reservoir and bringing it up to the houses. A few years an Italian NGO made a trip to Bungwe and installed an Italian water pump at the reservior which would pump water up to a tank in the village but soon that tank broke. The community of Bungwe reached out to EWB initially to fix the tank but after the UCSD chapter made a trip to the village, we decided that it was better to design a full fledged water catchment system for the village which will allow them to have access to water near their homes. The community was very happy with the decision to build water tanks and hence we started designing the water tanks.

After a few rounds of designing different types of water tanks, we found that individual house tanks which were small and could hold about 1000 gallons of water would be good because they were cheap and people could have multiple within their homes if they wanted. The community responded negatively to this idea stating that not many people could afford such tanks at the price point so we had think of something else. We decided to go with community tanks which were much larger in volume and would still be cost efficient. With the help of Chris Trautner, a PhD student in the structual engineering department, we were able to design the tanks and submitted the idea to the community. They were very happy with our progress but instead asked for $300,000 to build their own water tanks using the help of some local builders. The money would also be used to build a new school, a dormitory and a volleyball court. We definitely did not have access to that kind of money since we are a student chapter but we said that we would be more than happy to help them to build the water tanks. Since this has happened, communication with the community has been spotty but we have not given up. Currently we have put the project on hold and hope to return to it soon.

Since the recent developments with the community of Bungwe, we at EWB-UCSD are looking to apply for a local project that would be more feasable within a shorter period of time and I have noticed that domestic projects are a lot more successful due to it being much easier to communicate with the people since there is no language barrier. More people can also travel on these trips to assess or build since travel costs are a lot less.

While being the project lead, I have had to deal with a lot of things which have had a huge part of shaping me into the person I am today. First and foremost is learning with people changing their minds all the time and how to deal with that. EWB UCSD almost shut down a year ago due to a lot of students who were officers graduating and a lot of people leaving the club. It was me and a group of three other people who kept the club alive at the time. We worked together during the summer even though I was in India. Phone calls at 3 am (Indian time since I was in India for the summer interning at BGR) became a norm because we wanted the club to me more successful than it was in the previous year. I took on a lot of tasks that we had to get done. We got a new mentor over the summer who helped us with how to recruit a lot of members and build confidence in our team members. Once the school year came around, we had booked a room that could fit 60 people for the first GBM, and to our surprise 160 people showed up. Our hard work through the summer had paid off and we had gone from a club that had almost shut down 3 months ago to having such a turnout.

To attract more members, we had an intro project which was to build a water purifier which would cost less that $15. We split the students into teams and let them do their research and have team meetings to discuss ideas. At the end of 10 weeks, we let them build the filters that they had designed. This created a good bond amongst all the members that were joining our club and after this, they proceeded to help us with the design of the water tanks. During the school year, I lead a team of about 25 people in designing the water tanks and had to deal with all the things that came with leading such a large group of college students. I am now looking forward to leading another team of students in our next adventure with a new community.

This is how the community is currently collecting water during the rainy seasons and we are trying to improve what they are currently doing by building a water tank instead of all those small jerry-cans

This is the school in the community and behind the school is the soccer field with a goal built from wooden sticks and a make-shift volleyball court in the middle. We are trying to make all of these better for the community such that the kids would rather be out playing sports instead of doing bad things like joining gangs

As you can see in the picture above, they have no drinking water readily available in the community and this is the main problem we are trying to make this better

SD Hacks Project

This is the main part of my SD hacks project. I used HTML, CSS, JavaScript and Bootstrap to design this website. I learned all of these languages and mark-ups by myself during the summer.

I am also designing a website that will allow for hackers (students, professionals and who ever else is going to a hack-a-thon) to build a team before the hack-a-thon. I did not have a team before attending SD Hacks since it is my first hack-a-thon and was very unsure of what to expect. The website will allow for users to make an account, build their profile with that they have done at previous hack-a-thons that they have attended, the skills that the user has in terms of languages that he or she knows. Then the user can sign up for a hack-a-thon which will be displayed on the screen based on date and proximity to the user. Once the user has signed up for the hack-a-thon, he or she can view other users going to the hack-a-thon and see their profile. If the user wants to talk to another user who he or she is interested in making a team with, there will be a message option and also a team builder option such that it can accomodate multiple users.

Since at this point all I know are front-end languages, I have built the front-end for this website and will be building the back end either by learning the appropriate languages or joining a team at my next hack-a-thon for building this website. I will keep you updated on my progress with the website.

Ashwin Kannan

ABOUT ME

EDUCATION

ECE Coursework

Computer Science Coursework

Math Coursework

Business Coursework

Programming Languages