Maulik Shah

Courses

Projects

• Should I Donate (Winner @ GAI Appathon)
  

  Developed an Android App which would check the credibility of a Non-Profit by clicking a picture which would be reverse searched and their financial reports would be analysed to generate a rating. Used Java on Android Studio for the app, Google Image API for the reverse image search and OrgHunter API to get Non-Profit Details.

  Java , SQLite
• Ram Disk
  

  Developed a rudimentary version of the Linux File System on a partion of the RAM.

  C
• FIFO Scheduling Algorithm
  

  Emulated the FIFO Scheduling Algorithm on a dummy OS with pre-empting.

  C
• Keyboard Driver
  

  Developed a basic keyboard driver with provision for special keys on a Dummy OS.

  C
• Lung Cancer Detection
  

  We used CT scan slices to generate 3D models of the lung and using their HU values we generated a model of the coherent air cavity. We then trained a C-NN to identify contiguous voxels larger than a specific size which might indicate cancer. Our results were comparable to the Top 50 teams that participated in the Kaggle Competition. We used scipy, skimage, tensorflow packages to run the experiment.

  Python
• BU Campus Events App
  

  We developed an app for BU which would act as a common platform for all organizations to publish their events and for students to RSVP & Check-in to the event. Each oraganization can publish their events to the calendar with options like open-for-all or for selected majors and the respective students can see and access those events. The organizations can also issue ticekts and the students can check-in using QR codes. The backed and web-app was made using Ruby on Rails and the app was made using Java on Android Studio.

  Ruby on Rails, JAVA, PostgresSQL
• Analysis of StackOverflow Dataset
  

  Analysed StackOverflow Dataset from 2015 to find similarity between users based on the type of questions they answer and their scores using various Distance metrics. Also Performed a Time Series analysis on the users to find the similarity and dissimlarity between the users based on their activity time and to make deductions from that data.

  Python
• Smart Blender
  

  Built a smart blender using Arduino and Twilio API and a bunch of wood and plastic. User can text their order and the blender and the "machine" makes a smoothie for them and replies back when its ready. Lots of mess. Lots of fun!

  Arduino, JS
• Canteen Automation System
  

  Built a system in .Net and a supporting Android app. During lunch hours canteens get really busy and everyone cant be served efficently. So we developed a system where the students can place an order before hand and pick up his meal when he reaches the canteen. We also improved canteen efficiency by using a Forecasting model to predict demand for various dishes during various times of the day and dynamic inventory management which would be updated approximately with each order and would predict future state of the inventory and prompt the manager to restock various items. We also offered PoS order taking and printing and sales data visualization. We also performed Market Basket analysis for each user to generate personalized offers.

  .Net, Java, Android, Microsoft SQL
• Yelp Dataset Analysis
  

  Divided Las Vegas into culinary districts based on the tags and location of each restaurant using various clustering algorithms. In the Second part I realised some of the tags were wrong so I generated my own tags for each restaurant using NLP on the reviews for each restaurant.

  Python
• Photoshare
  

  Created a Flickr like website with a heavy focus on database design and efficiency using SQL and Flask on Python. Features included creating account, creating albums, adding photos to albums with tags, tag reccomendation based on previous activity, commenting on photos, adding friends, liking photos, showing top10 users, updating profile, etc.

  Python, Postgres SQL
• Modeling CO2 Emissions
  

  Given Countrywise CO2 emissions data combined with other statistics like HDI, GDP, Energy Concentration, etc we tried to model CO2 emissions and to predict a trend in CO2 emissions which works better than the Kaya Index.

  Python