Untitled

[jaredsmithse]

I would like to take some time to write about a scheme I executed last fall. The scheme was to get free rides from Uber or at the very least a very discounted rate. 

After a hackathon last summer/fall, I was talking shop with some teammates and one mentioned they heard of people running Google Ad campaigns to get referral credit from Uber. I thought this was crazy but brilliant, and later decided to give it a shot. 

I was in luck because in September Google ran an offer to give 75$ in free credit for spending $25 with a new account. Naturally, I opened 3 accounts under my different email addresses and now had $300 in Google Ad credit for an initial investment of $75. Over the next few months I managed to earn $600 in Uber credit.

In January, Uber raised its referral credit to $20 per referral, and I ran $20 for another campaign that yielded less people but did give me a net gain (I don't remember the specific amount). 

Uber only gives you the referral credit when users take their first ride, so I still have some people that signed up but have yet to take their first ride. Because of this, I still have referral credit coming in at a rate of about once a week. Some of the time delay I believe comes from the fact that when I did my first campaign I was targeting the whole US at first instead of only the cities that Uber operated in. 

Since that September of 2013 I have managed to pull in $970 in referral credit for my Ubers from an investment of $95. This number is lower than the actual amount because Uber does not record which ones got me $20 and assumes every one the 97 people on the 'accepted' list was only a $10 credit. There are still 28 people who have applied with my code but not yet taken their first ride, and every month the list gets shorter.

So, if you want to get some free rides from Uber while helping them grow their user base, feel free to follow in my footsteps and ride around town with the good feeling that it is at 90% less than the real cost.

[jaredsmithse]

HackSummit Post-Mortem

Data 2.0 hosted their first ever hackathon that was not part of a conference. They ran this one as a tester for a bigger one to take place later in the fall that is supposed to be amazing.

What sets them apart from other hackathons is that they have a social network for the event. When they first mentioned this and I looked at it online, it looked pretty dumb in my opinion. The site wasn't fully developed and looked like it was treated like some side project and not a main focus of the event. When they explained what their vision was for it though, I was impressed. 

They found that there were a lot of teams that would go to hackathons, build a product, then go to another hackathon the next weekend and build the same thing. This seemed like a terrible trend, and they were setting out to fix this with their social network. Once you make a profile on the social network you now have that project carry over into the next hackathon, allowing you to build on the functionality of the last project or just not repeat yourself without others knowing.

The theme of the project was to help empower companies or people to have a big impact on the Bay Area (or the world). My project idea was a simple app that would benefit my company specifically and hopefully help others). 

We at ZenPayroll try to go out for lunch every Wednesday, and we have found that as we grow, there are fewer and fewer restaurants that will take us, and the personal feeling of going out for lunch is often lost in the crowd. Our fix was to have a script with everyone's names that would shuffle us and assign us into groups that would then be assigned restaurants.

This was still pretty manual and didn't take into account whether people were going out (someone on support always needs to stay in), who was not in the office/on vacation that day, variety (making sure a group wasn't all engineers), and what restaurant each individual went to last week. Our app would do all this and if we were really lucky, it would also make the reservations with opentable as well.

My team consisted of myself and a classmate who just arrived for an internship at SmugMug for the summer, Andres. I had only competed with Andres once before at my first hackathon (NASA SpaceApps) and was really excited to see how far we had come and what we could accomplish.

Working for 24 hours at Change.org's HQ we managed to reach MVP with the exception that we did not put a physical button to text the attendees with options for the different restaurants (but we could do it from console), and we were not properly handling the responses from those texts.

Overall we now have a decent app at fieldtrip-it.herokuapp.com that you will be able to use after about another week of development. Ideally I would like to finish it for this Wednesday's lunch but realistically that will not happen. In a month I would like to have opentable/yelp and iCal integration as well for auto scheduling/replying/reserving. 

Our long-term goal would be to make this usable for all events, with the specific goal of being able to split my team into small groups and reserve times for community service at different locations (our favorite is GLIDE so thats the specific one we will shoot for) based on the availability in their calendars. 

Overall, I could not have asked for a better birthday weekend!

[jaredsmithse]

HackMIT 2013

A couple weekends ago I competed at HackMIT. I am really proud of what I did in the 24 hour window and hope to improve on it when I get back to SF or even before then if I can find the time. I originally went there with the intent of working on a team with someone I met from the a16z hackathon but he ended up having to leave halfway through so I decided to work alone. This was my first time going at it by myself with no one else to help me build a product, so even though I was sleep deprived (I stayed up for 43 hours that weekend and ~30 the day before with 8 hours sleep in between) I was excited to see what I could accomplish.

The idea I had was simple, a manager of a location where employees interact with customers needs a way of getting customer feedback without being there to watch the employees. The specific use case I had in mind was a restaurant (yeah I know it's the second restaurant themed idea this month...) where people are constantly checking in and leaving reviews, and these reviews are usually at the time of checking in so we have a timestamp. With this I wanted to have an app where managers can go on and put in there employees shifts and then the app would match up the checkin/review timestamp to those employee's who were on shift at the time. These reviews would undergo sentiment analysis using the AlchemyAPI to determine whether or not the review was positive, negative, or neutral. Additionally AlchemyAPI returns a list of keywords that have been isolated with an associated sentiment based on the context of the surrounding words, I stored these but did nothing with them during the time that I was there(more on this after).

The first problem I had was sleep deprivation. I had been up all night thursday before my flight (30 hours), and only had scattered sleep on the flight from San Francisco to Boston. That night I found myself at Harvard for a getogether my friend invited me to. I got back and didn't sleep until about 2am and had to be up and ready by 8:30am. That was the start of a 43 hour spree of being awake, working by myself (somehow I managed to stay up the entire night!). This led to a lot of slow progress on things I should've figured out or remembered much sooner.

One fatal mistake I made was two hours before submission. I had only then just remembered that Foursquare only responds to requests from https, and my computer could not do it (my environment has wierd issues with ssl certificates), I had no time to learn to use Iron.io (the solution to our last project that used Foursquare), and I figured that since I had other issues to deal with that I did know how to fix I decided to sacrifice live functionality. One thing that did not help me to remember was that using Foursquare's test keys you dont need https and I only moved to my own keys those few hours before submission.

A huge bonus to using AlchemyAPI was that there was someone there to represent the company as they were a partner for HackMIT. Whenever I had questions about how things were returned from the API or I was having some kind of problem at all interacting with it, I would run right up the bleachers to the guy and he was more than happy to help.

Aside from AlchemyAPI, a cool tool I had previously not used before but decided to incorporate into my project was D3.js. I have been meaning to work on my javascript skills and thought that this would be a good chance. I had one chance to use this before at the NASA SpaceApps Challenge last spring but one of the two people on my team was really good with R so we built a webapp using R/Shiny which made D3.js uneccessary for us. I didn't get around to implementing it until near the end of the competition so I did a little bit of copy/pasting from an example directly into the template view (not my proudest moment...) and worked around that to get the data to show correctly and nicely for each employee (I used colorful donut pie charts).

Usually I am excited to write up a description for the apps I make at hackathons, but whether or not I consider the different challenges I faced, I consider my app to be incomplete. It does not have the polish of others that we have made in the past and I think that if I put some more time into it I will be able to put it up for people to see (and on my LinkedIn), but as any engineer excited to create new things, I already have two new ideas to work on when I get back to SF (along with my contract work and the job hunt) so it will be difficult as always.

[jaredsmithse]

Emirates Travel Hackathon

I was supposed to post this at the beginning of the week, but now I have put it off until today and should post now before HackMIT is done. I originally planned on participating in three hacking competitions starting with HackMIT, but now it looks like I will only do two. Either way, I wanted to write about my experience at my last hackathon so that I can start to have some practice with blogging and so that I can write down everything I am doing at these events.

Since this was Emirates' Hackathon it is expected to be travel themed. We (team of 5) built a product (called 'TCKT', with a domain at tckt.es) that will allow you to input and keep track of your itinerary online and will match your itinerary to other's to offer suggestions of what you could do to fill in the gaps in your itinerary. This allows people who do not know the area they are traveling too well and would like tailored recommendations of what they could do while there.

There were a couple of things that we knew was a stretch but wanted to implement if we could. They are as follows: - Voting/ranking system for recommended events(to have a way of ordering the recommendations in big cities) - Eventbrite and Xola integration (Xola is only in SF but they were there offering a prize for using their api) - Tumblr integration for blogging about each event - Photo upload for events

We ran into a couple of problems along the way, most of them team based (large team of talented developers and little communication). Our biggest problem by far was the large team. We had five fairly talented Rails developers that were not good at managing each other. If someone was told to do something they would get it done much faster than expected and start working on something else. This led to a lot of overlapping work that could have been avoided with proper project tracking tools like pivotal (or taking the trello board seriously). Another big problem was communication, which kind of ties in with the first problem. We all came from the same camp, so we all know how to work together, and whenever one of us got stuck we would go to another teammate for help. this was great, but whenever one of us went to help out we would not always complete an older task. For example, I was building up the algorithm for finding recommendations but was called on to figure out how to figure out a way of getting around the google api problem (more on that next next) and I didn't get around to making the correct routes for recommendations until the next day while someone had already worked on it.

Working at Microsoft's had a unique challenge as we were banned from Google api calls. We're not sure if this was because many teams were using Google's apis or just because we were working from a Microsoft IP address (we were one of the first teams to start working and ran into the problem almost immediately). I was tasked with finding a solution and tried several things (even SSHing into one of the school servers at RIT in NY to try and call from a different IP address but they wouldn't let me install gems and loading them was not working correctly for some reason when loading manually worked on my local machine) before finding geocodefarm.com which allowed me to make manual POST requests to their site which I assume was just a proxy to Google maps api. Some tradeoffs were we wasted valuable time finding this solution, the reverse coordinate lookup function brought back inconsistent address formats, and there was a small limit of 100 requests per day so we had to switch accounts by the time presentations came around.

Mobile design is hard for a bunch of engineers that have never done it in any form of a production scenario. We had our two most willing CSS guys dedicated to the mobile design for almost the entire time of the event. They did some things of CSS and SCSS they were not proud of, but they managed to make a great looking site for phones which was key for our travel-themed project (we had no experience developing native mobile apps). We wanted photo uploads to work but only got them to partially. We were having trouble using paperclip and I ended up taking their demo app and dissecting/comparing/building their code against ours piece by piece until it worked. The fatal mistake I made was that I didn't test it on Heroku until hours before we were done and found that it worked locally because Paperclip was saving the photos locally in the project and not to AWS S3 like we thought, which was a problem because Heroku does not allow local storage like that and we did not have the time to fix it. We ended up with is a Rails app that only partially integrated the photo uploads. The site is also mobile responsive and works functionally with a few bugs (we got mvp). Because of our poor management skills over each other we did not get to reach a lot of our stretches. We ended up making it into the final 20 (out of 60) that did a second round of presentations up on stage to the entire crew at the event but we did not get any of the top three awards.

One of the teammates suggested we build a framework like rails-composer that sets a bunch of boiler plate code up to help us ship an mvp faster at future events. One of our teammates had never done user auth before and we decided to have him do it so that he could learn, but the consequence was he did not use something like devise, and did not roll his own encryption scheme either (don't use your real password when/if you sign up). It took a lot more time for him to get it right even though one of the other teammates could have done it quicker, but it was important that he learned something that was outside of his comfort zone. We also all unanimously agreed that with bigger groups we unfortunately need more overhead to be able to have maximum productivity whether we like it or not.

You can have a look at the site at tckt.es and the code is located at https://github.com/NathanielWroblewski/tickets if your interested in taking a closer look!

Keep checking back for updates on the next three hackathons as I compete at HackMIT, LincolnLabs Education Hackathon, and HackUpstate. These will all be with varying team sizes and from various locations.

[jaredsmithse]

NASA SpaceApps Challenge

This past weekend I participated in the NASA SpaceApps challenge with two classmates, Andres and Brian. We chose the Earth Day Challenge as part of the Data Visualization category. I'm so late writing this because right after I slept for 15 hours, woke up, and studied for two tests the next day on Tuesday. After my tests, I studied for another exam I had today.

Today I am relaxing. I'll do some programming problems on Hackerrank or Project Euler and work on the PortalFace Project for my web engineering class, but I just wanted to share my experience at the competition before it is no longer fresh in my mind (arguably already too late).

I have never competed in any sort of coding competition in my life, I never thought that I was good enough to really contribute to a team let alone go it alone but Andres convinced me I should work with him and two friends at NASA's challenge (one of them was a designer who did not show up, I ended up having to do his work as well).

NASA had five categories (Citizen Science, Creative, Software, Hardware, Data Visualization), and we met before the competition to look over the ones we thought would be the coolest to work on. We decided on a remote LEGO robot control system, renewable energy potential mapping, making a near earth object database, and our earth day challenge.

The challenge was to show how the earth has changed over time with things like air quality and climate change. Our original idea was to have a globe you could spin that would have a slider and drop-down to select the data you want. When we examined it realistically we decided to start with just the US land area and then if we had time move on from there. This turned out to be a really good decision later on.

The first SIX hours were spent looking just for data. Government and really any form of public database is extremely hard to navigate, and even when you find the data you have to get access to it. I ended up finding the primary data used in our project (NOAA temperature, rain, and snow), but they took forever to get it to us (we had to submit a request and did not get access until 4 hours before time was up). Even when we got this data we had to write scripts to format the data for our site.

The site would be built entirely on an R framework called Shiny. We knew we would be using R for the data visualization itself but when Brian (he worked a lot in R on his co-op at Google so he was our expert) mentioned there was a web app framework we could use we were ecstatic, because this would set up our work structure as Andres and I mining and formatting data to throw over to Brian to format while he set up the framework. After we got data we would then move to building the rest of the site (it had bootstrap built in!) while helping Brian the whole time (he hardly needed it so this was good).

We ended up staying up the entire 26 hours, and learned many things. Most importantly we learned there was a reason why no one else competed in our category locally (manipulating data is hard). We naturally took first place in our category, with a partially finished app that accomplished a lot but had a ways to go. We managed to get temperature, rainfall, and snowfall by county working from 1990 to present but fell short of getting air quality (I got the data but we did not have time to include it). We were also supposed to display data from 1970-present but when we tried to run the 4,082,000 lines of data through R with a quad-core, 16gb RAM, and 40GB swap, we had to kill the program or crash the computer (which we did not want since it was running our server and we were working on it remotely).

The app looks decent but is not complete, especially since we did not find an online hosting solution. You can check out the project page at climatetimemachine.com which will also have a link to the github repo for our code. 

I suprised myself at what I could contribute to a team in such a fast-paced environment, and I will now start to actively seek out more experiences like this in the future.

[jaredsmithse]

Using Ontology for Requirements Elicitation

This last week I was given an assignment for our Requirements and Architecture Engineering class to look up a new technology or method of going about requirements or architecture engineering and then present it to the class. I didn't really know what to pick since I didn't feel anything would be very interesting to me, but a search of 'requirements engineering' in our library's database returned an article for enhancing performance-based requirements elicitation of financial systems using ontology.

Reading into it I decided that using ontology in general for elicitation was the most interesting part of the article so I researched it more and made this my main topic. Ontology is essentially (for the purposes of this post) the study of information, and it's approaches to requirements elicitation are very helpful. Normally for elicitation you would use several methods, with the most popular being interviewing. You normally have a set of questions beforehand and you ask the client new questions on the spot as well to try and cover everything you think you need to make the software for them (usually you record as well). A popular question that always gets asked towards the end of the interview is "Are there any other questions we should be asking?". A technique I like is shadowing. You often get to 'intern' with the people who will be using the system (if it is for a company to use internally), which gives great insight on what works, what doesn't, and a good idea of how the system will be used on a day-to-day basis.

Ontology is not meant to replace methods of elicitation like these, but to make sure that you did not overlook anything. What you do is take the information you have on what the system will do and model it so that a model checker can look for errors in your knowledge-base. For the example of the financial system you have a list of use cases outlining the general requirements of the system, you know the non-functional requirements (things like the security and performance of the system), and in the case of this application domain, you have a lot of laws and federal regulations to abide by. With a modelling language (the most popular one is OWL), you will construct the system and place all the constraints into it. The program will then run and tell you if there are contradictions or specific instances where the system will do something that you didn't want it to do. Depending on the problems that arise from the model, it will give you a good idea on if your desing is flawed, if you forgot something, or it can even show things you didn't know were a problem yet. From there you can change/improve your requirements and keep iterating these steps and running the model to get everything right.

This idea of iterative development of your requirements for a software system is very powerful, and tends to give you a confident feeling that you are headed in the right direction much like Test-Driven Development does. In this case, however, you are catching problems before you even get to coding which lowers the cost of fixing faults in your system dramatically