Joshua Mazariegos

UoFT STUDENT HELPER WEB APP

JAVA, SPRING BOOT, MONGODB, HTML, JAVASCRIPT, CSS, SQL

Github Repo (Currently Private)

What is it?

An app built for a class with fellow classmates that was built with the intent of helping other UTM students with features such as a calendar, note taking, friends, timetable, and a map of the campus with food locations and times
This was built using mySQL, mongoDB, Spring Boot, HTML, CSS, Java, jQuery, and Javascript along with making use of REST endpoints to have interaction between our databases and client which was used for the timetable for courses, calendar for dates, and for user information
Calendar lets you create dates with different importance like due dates, regular events, etc. along with automatically placing class times on the calendar if enrolled in the timetable
Note taking allows taking notes as the name suggests but also allows you to save them for later use, and then load them up again if needed
Timetable lets you enrol and unenrol in courses scraped from the UTM Timetable website with courses being related to a certain session and semester so you will only see courses from the respective session and semester, you are also able to upload your own timetable from the UofT website, and lastly are able to have a cart of courses that you can then generate a timetable with smallest amount of gaps between courses
Map allows you to look at a map of the campus and see what respective food places are at each location and their times it also is green if the store is open

What did I do?

I worked on the timetable portion of the project mainly, with only a bit of work on some of the javascript for the calendar
I worked on a scraping python script to scrape the courses from the UTM Timetable website and then place into our mongoDB database for later usage in the timetable
I also worked on the HTML, CSS, and Javascript for the client side of the timetable which is using jQuery to POST or GET the information needed and then using that information to do things such as display courses available for that session and semester, display a user's timetable and enrol and unenrol from courses
More on the client side I did is the timetable generation cart section, so you are able to add to cart and remove from cart which courses you would like to generate then send a request to generate your timetable which then auto enrols the generated times (I did not write the timetable generator part)
On the server side I worked on the endpoints such as displaying the courses which was written all in Java using Spring Boot linked to our mongoDB database, and also the enroling / unenroliing server side so what controls a user's timetable and displaying the timetable

DEVICE FINGERPRINTING API

REACT, SQLite, JAVASCRIPT, EXPRESS.JS, NODE.JS

Gitlab Repo (Currently Private)

What is it?

A device fingerprinting authenticator API, built for a project course with fellow classmates with the intent of testing device fingerprinting as a form of second factor authenticator where it is hassle free, and almost unnoticeable by the user
This was built using SQLite, React Native, Yarn, express.js, Javascript, HTML, CSS, and Nodemailer where we use React Native to create a homepage where you could register or log in, and then we have Yarn to create our server where the API works, using REST endpoints through express.js to do certain things such as logging in, registering, and certain database queries through SQLite for example determining devices tied to an account, and also have scripts to send mail to emails for confirmation and also to confirm identity through unique codes
When registering it would send you an email to confirm your registration and also run the fingerprinting script to register a computer with its fingerprints so then you could check the fingerprints when you log in.
When logging in it would run the fingerprinting script and confirm that your fingerprints are similar with a threshold of 90% and if under the threshold would log the new fingerprints that didn't match and email you to confirm that this attempt was you so then it can register a new computer with new fingerprints to your account so that you can use it to log in without removing the other computer and its fingerprints
After logging in, you have the ability to remove devices tied to your account and also see which accounts are tied to you through database queries using SQLite and express.js along with fingerprints we collect, so we can see the fingerprints that can be passed through

What did I do?

I worked on the backend portion of the project so I worked with the database and the endpoints with a little bit on the React Native end for sending data through the client to the server
For the database and endpoints I worked on the queries and relationships, such as creating a relationship for users, devices, fingerprints, and tokens then having queries so given a user I could get a list of devices or create a expirable token for users so they can confirm their identity through their email
Along with that I worked on the script for sending and email and creating a token for when you attempt to log in from another device, which was using nodemailer and crypto, nodemailer was used to send the email through a gmail account that we owned, and using crypto to generate a unique token which then gets registered in our database to a user with a expire time and if they can put their token in before it expires it will do device registering and all

MINESWEEPER AI

HTML, JAVASCRIPT, CSS

What is it?

A minesweeper game with an AI attached to it!
The AI is meant to learn from losses in 3x3 squares and use that knowledge to try to win, only leverage it has is basic logic i.e. 3x3 with 1 square and only 1s meaning that has to be a bomb

What did I do?

I created the entire minesweeper game from scratch in javascript along with the AI, the visuals were all created by me as well
The game was done using clickable images where the board is created by the createBoard function which places the amount of tiles specified and then randomly places the amount of bombs for the game but avoiding where it was clicked to avoid losing immediately and the buttons when clicked to a recursive function to check whether the next tiles around are empty and if they are continue checking around them until they hit a tile with a bomb near it which then marks that tile with the number of bombs near it
The timer is done using a timer interval same with the placing the flag down with left click but can also be done instantly with right click, technically you can make a board of any size but for limiting (and other reasons) I have limited it to the normal amount, the reset button is the face and will always create the board with the settings that are there
The AI on the other hand was made with a reinforcement learning method in mind so its a 3 layer AI with the input layer as the full board, the hidden layer as the 3x3 squares that are on the board, and the output layer as the single tile from the 3x3 square, and the hidden layer is made up of only one version of each 3x3 square so that each board influences the probability of a 3x3 square by making it easier to determine its value, routing to output, and smaller size for AI
When the AI starts it adds the current board into the network then adds all new 3x3 squares within the board and attaches any existing 3x3 squares in the network to the board with the corresponding positions of each tile in the 3x3 square to the full board (real position on board) and the connection from the 3x3 square to the tile containing the position of the tile in 3x3 square (relative position)
The probabilities start at an equal level for each tile that isn't revealed yet so a new 3x3 square with 3 unrevealed tiles will have each tile with a probability of 1/3 of being a bomb and depending on what the percentage reward is, it will increase tiles that have not been a bomb (while decreasing the others), or decrease them (and vice versa) which happens through a backward pass if the tile is a bomb or not and the forward pass finds the tile with the highest probability
Lastly the AI has settings for making it do runs on its own so it can get trained and the delay if you want to watch it, which is still in development for the time being
Bonus! I noticed that the AI was really bad at learning with no prior knowledge so I attempted to give it a little leverage by giving it some basic logic of where bombs MUST be so spots where it has to be, it will already know by using a function and using the faces of the tiles already seen

SELF DRIVING CAR

PYTHON, ROSPY, FLASK, NVIDIA LEARNING NETWORK, DASH

What is it?

This is a research project I did with a professor and my classmates in my final year, the goal of the project was to create a self-driving car then create scenarios where the car would not be able to properly make a decision in order to show potential flaws within self-driving technology that can then be used to help avoid those scenarios.
We were able to create the self-driving car using a remote car equipped with a liDAR, RGBD camera, and a small computer running Ubuntu, it first begun using a PID Control to create the dataset which would be used to train the neural network while it worked for our room, in our campus, the building we were using had glass fence and windows in the hallways causing the liDAR to be unable to guess how far the wall actually was, making the dataset more unreliable, so we switched to using a controller and driving it in the way we would want it to drive.
Afterwards we would trained the NVIDIA network based on our dataset to have the car create the decisions, so we had to re-weigh some values such as turning since compared to the going straight data we had, we needed it to properly weigh the turning data so turning is prioritized during turns rather than going straight.
In order to see it moving live and decision making as it was driving, we had implemented a DASH/Flask website hosted on the computer on the remote car which would store and create live graphs of the angle and time data so when it would make a strange decision we could see when and why it happened, helping us work on it faster along with a remote command terminal to control the car's server and network.
As a sad final note, the pandemic had started during the final months of our research project so we weren't able to work on the create scenarios where our car would fail because of the school being moved to remote learning and no students allowed on the campus unless you live there, making it much harder to work on it.

What did I do?

I helped with working on the neural network, creating the DASH/Flask server, and a few experimental testing choices.
Regarding the neural network what I worked on was helping the decision making and weighing of the network because when we realized that the data heavily prioritized going straight because for a majority of the data the car is going straight so turning left and right was weighted less, so I helped in determining how to adjust our weight function so that we would correctly weigh the data in order for our remote car to properly drive through our campus building's hallways.
In regards with the DASH/Flask server, I created the server using Flask and REST then using DASH to create the graphs which would be updated live and recorded so that we could review the data as it happens or if we need it later.
I used REST in order for the website to receive live data, so it would send the time of decision and angle of the car which would then be input into the current CSV file for the run then have DASH update based on the CSV we just updated using a DASH callback function, while also including tools such as pausing the graph update in case we need to look at a specific time while it stays updating or resetting the graph for new data / run.
Final use for the FLASK server was a remote terminal just using REST, SYSTEM, and a terminal sandbox iframe made by ethanhuang0526 which would take your command from the terminal then send it to the server on the remote car to then execute commands such as starting the network or working on it as it is driving, as before that we could only work on it when we hooked the car up to the monitor in our room.
Continuing with the decision making and experimental testing, I helped to implement PyTorch's floor segmentation to then feed the car only data of the floor so then the car would only need the floor to know what decision to make, and we were able to segment the floor from test data images but unfortunately we weren't able to put the data to the test as pandemic had started making using the remote car more difficult.

THIS WEBSITE!

HTML, CSS, JAVASCRIPT, JQUERY

Github Repo

What is it?

My website, you may ask why since you can just make one easily anywhere and to that I say I like learning and trying out new things, something nice about seeing something you make come together!
Maybe more coming eventually!

What did I do?

Javascript + JQuery to change pages using a function based on where the page is located and which page will be changed to using JQuery for easier element changes such as disabling all clickable buttons / dropdowns and using them to change animation classes for page changes
CSS is used to do the animations and are tied to classes so that when the page changes the relevant elements start doing animation by having animations for fading out and fading in mixed with the timeout function to make sure the animations look smooth and everything is gone before replacing

PDF VERSION

WORK EXPERIENCE

TATA CONSULTANCY SERVICES - QA ANALYST

APRIL 2021 - PRESENT

Coordinated with a team, and was a task coordinator for my team including leading, assigning tests, and creating test cases.
Worked to help team members if they needed help with software, test cases, and documentation.
Trained and helped new team members who joined the team through on call and hands on, making sure they were comfortable with the work.
Tested and created tests for software products and bugs within a timeline and coordinated with team members to complete tests.
Created test documents and emails based on analysis from existing test documents and test cases.

PROJECTS

UoFT STUDENT HELPER WEB APP

JAVA, SPRING BOOT, MONGODB, HTML, JAVASCRIPT, CSS, SQL

Created a webpage with classmates for the timetable and course search using HTML, CSS, and Javascript.
Displaying courses chosen in the timeline on the big calendar.
Created RESTful API endpoints for getting and posting data from our database in JSON format, then having the data interpreted used for enrolling, searching, generating, and maintaining the timetable.
Uses scraped data for the endpoints which are implemented using mongoDB and Spring Boot.
Set up the scraped data in a nice JSON format to be read in Java Spring Boot and storing individual user's timetables

MINESWEEPER AI

HTML, JAVASCRIPT, CSS

Created a Minesweeper game to use to train a network I created to see if it can learn every unambiguous possibility.
Minesweeper game built using just Javascript which is capable of being any size with any amount of bombs, and being able to flag or reset.
The game also uses timers and onupdates to keep track of bombs, flags, and time taken for the current run.
The network is 1-level neural network which learns through reinforced learning to learn which spots most commonly are bombs or not.
Uses just the knowledge of the numbers in the square to create decisions, and can be set to play a set amount of games.

SELF DRIVING CAR

PYTHON, ROSPY, FLASK, NVIDIA LEARNING NETWORK, DASH

Worked with classmates to make a remote car be able to drive through our school hallways on their own.
Created a PID Controller using liDAR on the car so it would be able to follow walls at a certain distance.
Collected photos and respective speed / angle using a camera while we controlled it to make the training data and test data which would then be trained in pytorch using a Nvidia Network.
Used a Flask/DASH server to record the car’s angles as it drives to see live how it’s driving for any outlier outputs.
Created a command terminal on the Flask server, so we would be able to work and use commands on the car as it was driving.

DEVICE FINGERPRINTING API

REACT, SQLite, JAVASCRIPT, EXPRESS.JS, NODE.JS

Working on interpreting all the fingerprint data that gets sent to the database and storing it away linked to a user.
Worked on a email two factor authentication which when you can’t be authenticated using your fingerprint, authenticates using an email and updates your fingerprint.
Worked on an authenticator to use fingerprints to authorize access using API endpoints which can be used on any website through our server along with device lookup and deleting.

JOSHUA MAZARIEGOS

UoFT STUDENT HELPER WEB APP

JAVA, SPRING BOOT, MONGODB, HTML, JAVASCRIPT, CSS, SQL

What is it?

What did I do?

DEVICE FINGERPRINTING API

REACT, SQLite, JAVASCRIPT, EXPRESS.JS, NODE.JS

What is it?

What did I do?

MINESWEEPER AI

HTML, JAVASCRIPT, CSS

What is it?

What did I do?

SELF DRIVING CAR

PYTHON, ROSPY, FLASK, NVIDIA LEARNING NETWORK, DASH

What is it?

What did I do?

THIS WEBSITE!

HTML, CSS, JAVASCRIPT, JQUERY

What is it?

What did I do?

WORK EXPERIENCE

TATA CONSULTANCY SERVICES - QA ANALYST

APRIL 2021 - PRESENT

PROJECTS

UoFT STUDENT HELPER WEB APP

JAVA, SPRING BOOT, MONGODB, HTML, JAVASCRIPT, CSS, SQL

MINESWEEPER AI

HTML, JAVASCRIPT, CSS

SELF DRIVING CAR

PYTHON, ROSPY, FLASK, NVIDIA LEARNING NETWORK, DASH

DEVICE FINGERPRINTING API

REACT, SQLite, JAVASCRIPT, EXPRESS.JS, NODE.JS

TOSTADAS

FLAUTAS

CARNE SOUP

ENCHILADAS

PUPUSAS

SQUASH SOUP

SPAGHETTI

LASAGNA

PIZZA

POUTINE

FLAN

TRES LECHES