I implemented this algorithm (paper) which measures user credibility in social media by comparing people's tweets in time intervals and then clustering users to detect suspicious and negative coordination. The most cumbersome part was creating a cosine similarity matrix for the FakeNewsNet dataset's users due to not having a GPU or CPU cluster. It was done manually by over 150 Google Colab notebooks simultaneously in 2 weeks!

• github.com/imtore/CredRank-backup 

We implemented a food order website called "Loghme"  (Java EE Tomcat + Spring WebMVC, React, MySQL) for this course and then deployed it utilizing Docker-Kubernetes. In this web application, you could:

1. order food

2. search for foods and restaurants

3. pay

4. add to your credit

5. buy near ending foods from restaurants every 20 minutes with discounts (Food Party feature) 

6. the delivery assignment was simulated and you could see how much longer does the order take to be delivered

7. login, signup, signout (also with your google account)

8. see your previous orders

9. you also had a shopping cart

• Loghme Backend 

• Loghme Frontend 

• Spam Detection with Naive-Bayes 

• Breakthrough Game and Alpha-Beta Pruning 

• BFS, Iterative Deepening and A* for 8 Queens 

• Testing different classification algorithms on MNIST 

This was a compiler for Smoola language (Smoola Documentation).

• In Phase 1, the lexer of Smoola is made in Smoola.g4 using Antlr. In Phase 2, the AST is constructed from lexical analysis and the symbol table is filled. Finally, it shows up semantic errors. Phase 1+2 

• In Phase 3, the remaining errors are shown and type-checking is done. Phase 3

• In Phase 4, Java bytecode of the Smoola program will be generated. Phase 4 

Worked in a team of five, to develop a platform like Google Drive for students of the University of Tehran, with the following frameworks/languages/programs: Angular, Kotlin, Spring Boot, MongoDB. We then deployed it on a server to add download features.

• Backend 

• Frontend