Overview

A database solution designed to help veterinary clinics manage pet records, owner information, and appointments seamlessly through a GUI.

Technologies

Features

• Pet and owner record management
• Appointment scheduling system
• Medical history tracking
• Search and filter capabilities
• Data validation and error handling

Development Process

1. Designed database schema with tables
2. Implemented primary/foreign key constraints
3. Created stored procedures for data entry
4. Established entity relationships
5. Developed scalable database structure

Overview

Chalkboard is an online interactive canvas for remote collaboration, where users can draw, add objects, and create session-based workspaces that instantly sync across web browsers and all devices. Each session has its own unique canvas, allowing multiple groups to collaborate simultaneously within their own dedicated spaces.

Technologies

Features

• Drawing features (pen, drag, eraser, color change, shapes, undo, etc.)
• Real-time collaboration
• Login/Signup using Google Auth
• Load images or PDF
• Create/Join sessions
• Save updated canvases snapshots in real-time
• Collaborate with multiple users via video/audio call or chatbox
• Picture in picture (PiP) to view other applications while drawing
• Save/Load canvases
• Light/Dark mode
• Cross-platform compatibility

Development Process

1. Documented requirement specification
2. Designed multiple diagrams (UML, Use Case, etc.)
3. Assigned work into multiple phases through Rally
4. Developed features on separate branches
5. Tested the application and fixed bugs
6. Integrated features to the main branch
7. Ensured the Rally had a proper burn-down chart each week
8. Deployed the application on Render
9. Documented the test procedures and test cases
10. Presented at Temple's Symposium after being selected

Overview

Story Spinner is a website that allows you to generate a story given some text or using a random seed. It uses NLP and Word2Vec to generate and summarize the text to create a story

Technologies

Features

• Trained on 9 different datasets
• Generate a story given text or use a randomized seed
• Word2Vec for text summarization

Development Process

1. Created rough drafts using Use Cases and UML diagrams
2. Divided the work among 3 members
3. Gathered datasets of various topics
4. Trained the datasets and saved the models
5. Implemented Word2Vec to generate text and summarize

Overview

This project was made to address a infestation problem that was happening at a classmates job. Instead of having mice harmed and trapped, we designed a solution to bait and trap them so they can be released later on.

Technologies

Features

• Sensors to detect mice proximity
• Servo to open and close the door
• Room for improvement such as detecting box temperature

Development Process

1. Learn how to breadboard
2. Setting up the breadboard to use sensor and servo
3. Implement the code to trigger the servo when mouse is detected
4. Fine tuning the sensor to detect how close the mice needs to be to trigger the servo

Overview

Celebrity Look-alike is a fun and engaging web app that uses Clarifai's pretrained celebrities model to detect who you most likely resemble based on facial features. It was trained on over 10,000 celebrities across various categories, including actors, athletes, musicians, and more. Users upload an image url link, and the app detects the closest match among the preloaded celebrity dataset.

Technologies

Features

• Login and Registration
• Face detection
• Facial comparison
• 10,000+ celebrity database
• Handling users via database
• Hosting backend and frontend on Render

Development Process

1. Found a pre-trained model that was trained on as many celebrities as possible
2. Tested by passing an image through an url link
3. Created a sign/signout feature to have some sort of rate limit
4. Created a database to store user information
5. Implemented the website to allow users to pass in a link after creating an account
6. Display the top result
7. Tested on a few celebrities to assure accuracy (example: Bill Gates)
8. Hosted both backend and frontend on render

Overview

This research project developed a neural network model to detect and classify various eye diseases from retinal images with high accuracy, aiming to assist medical professionals in early diagnosis.

Technologies

Features

• 3 eye diseases and normal
• Trained on 85000+ images
• High-accuracy detection
• Different angles of each disease

Development Process

1. Import common libraries
2. Used Google Colab for faster GPU
3. Setup GPU
4. Getting the data from my Professor
5. Build/Train a CNN
6. Improve the model using pre-trained models such as VGG16, Xception, and MobileNet
7. Improved model further with hyperparameter tuning
8. Save the model
9. Present the results visually

Overview

The Color-Proportions project is designed to analyze color proportions in video or image frames and plotting it multiple types of data graphs

Technologies

Features

• Video/Image processing
• Extracting frames from videos
• Sending color analysis from MATLAB to Java
• Plotting different types of graphs
• Java GUI to display the results
• Java and Arduino communication (using RXTX library) to point to most dominant color

Development Process

1. Defined project goal, assigned team roles, and planned the workflow
2. Developed MATLAB scripts to analyze color proportions in image and video frames
3. Exported MATLAB analysis results as figures and text files
4. Built a Java GUI to display results and handle user file selection
5. Implemented file transfer and data sharing between MATLAB and Java
6. Programmed Arduino to control a servo motor based on dominant color data
7. Established serial communication between Java and Arduino (using RXTX library)
8. Tested and debugged the complete workflow from analysis to hardware response
9. Documented the process and created UML diagrams for project clarity

Overview

Developed an AI email responder using LLMs, RAG, LangChain, and ChromaDB for one of the world’s largest shipping companies. The email responder uses RAG to answer questions about package issues and general information.

Technologies

Features

• Email response generation
• General information
• Shipping statues
• Transit statues
• Delivery status
• Package delays (ex: investigations)

Development Process

1. Store previous emails in a CSV file with labels
2. Create chunks of the emails an embed them using OpenAI embeddings
3. Store the embeddings in ChromaDB
4. Take user query and embed it using OpenAI embeddings
5. Perform a vector similarity search on ChromaDB to find the most relevant emails
6. Use LangChain to chain the RAG and the LLM for a customized email response based on the user's query
7. Generate the email response using the LLM
8. Developed Azure-integrated APIs for metadata retrieval and package tracking

Overview

This project involved making decisions around scalability, cost, and performance while configuring GCP services. Migrated billions of records from Oracle to BigQuery, assiting with multiple GCP tools, and implemented a solution to compare billions of records for data validation.

Technologies

Features

• Python script to validate 1+ billion data
• GCP infrastructure support
• Striim to offer real-time data transfer
• GCP Certified

Development Process

1. Define where the data is coming from (our case Oracle) and Striim will use adapters to connect and capture changes in data
2. Incoming data flows through a real-time pipeline
3. Implemented efficient data pipelines
4. Process the data written in TQL (Striims SQL equivalence)
5. Processed data is sent over to BigQuery
6. Monitoring with Striim to ensure the health of the data transfer
7. Using a Python script to validate that all records of Oracle match with BigQuery
8. Throughout the project - assisted with architectural decisions around scalability, cost, and performance while configuring GCP services