Visaj Shah

1. Developed a full-stack stock trading application using Java and Spring with a data persistence layer built using JPA and Hibernate. PostgreSQL serves as the database.

2. The project was presented to Intel Corporation Leadership upon successful completion.

k-means Reimagined for Graphs

1. Wordle with a twist for cinephiles. Instead of guessing randomly, players guess a film title using keyphrase hints provided in addition to the color hints.

2. Designed using React.js hooks and components and a JSON database for movie titles.

3. Application deployed on Heroku with Google Analytics enabled. Reached 50+ players in less than a month.

1. Identified community clusters in a large network dataset using Graph Neural Networks (GNN) and
Gumbel Softmax method.

2. Extended the algorithm to allow for multiple clustering based on custom probability threshold.

3. Proposed an iterative method based on modularity maximization to determine the ideal number of
clusters in a complex network.

Abstract:
Options pricing has always been an important mathematical problem in Quantitative Finance. Among the traditional models, the Black-Scholes-Metron (BSM) model was considered as one of the biggest breakthroughs. While the BSM model is widely popular, and has an appreciable accuracy, it makes certain assumptions. These assumptions limit the performance of such models, and can be overcome using the modern solutions. With time, as the computational capabilities increased, and… Show more

Abstract:
Options pricing has always been an important mathematical problem in Quantitative Finance. Among the traditional models, the Black-Scholes-Metron (BSM) model was considered as one of the biggest breakthroughs. While the BSM model is widely popular, and has an appreciable accuracy, it makes certain assumptions. These assumptions limit the performance of such models, and can be overcome using the modern solutions. With time, as the computational capabilities increased, and researchers set out to address the limitations of the BSM model and other traditional models, Machine Learning (ML), and now Deep Learning (DL), have been increasingly used to develop better options pricing models. This is a relatively young domain, and there is a lot of scope in the field. ML/DL solutions leverage the potential of historical data in identifying patterns in price movement. Such models are quite accurate at detecting non-linearities in price variation, which is very important in increasing the accuracy of the predictions. Further, after developing the ML/DL models, we will compare their performances with the BSM model to evaluate the differences in error and accuracy. The historical dataset preparation will include data extraction from NSE India site and pre-processing the data on our end. We will have to write functions to find values of some important parameters, which are required by our models. Show less

1. A desktop application that sends reminders at regular intervals to the user to take a short break and relax.

2. Project Manager leading a team of 10 members; oversaw the entire Agile SDLC from SRS to testing.

3. Developed front-end pages using HTML5, CSS3, and JavaScript based on wireframes designed in Figma.

4. Performed Black-box and GUI testing using tools like DeepSource.

Problem Statement:
There are vast applications of Internet of Things and Machine Learning in the field of Aquaculture. One of the most critical factors in ensuring a healthy and productive livestock is the water suitability to sustain life. With this project, we aim to apply machine learning algorithms to data of parameters of water, mainly, Temperature, pH Level and Turbidity. We study these characteristics and develop a model to predict the suitability based on the data… Show more

Problem Statement:
There are vast applications of Internet of Things and Machine Learning in the field of Aquaculture. One of the most critical factors in ensuring a healthy and productive livestock is the water suitability to sustain life. With this project, we aim to apply machine learning algorithms to data of parameters of water, mainly, Temperature, pH Level and Turbidity. We study these characteristics and develop a model to predict the suitability based on the data available.

Further details can be found in the Project Report: https://github.com/Aayush-Desai/iot/blob/main/report_201801016_201801060.pdf. Show less

1. Worked in a team of 4 to create a SQL-based database in PostgreSQL for blood banks to manage data related to inventory, donors, patients, camps, staff, and transactions.

2. Successfully executed the entire development process from scratch including Requirements Elicitation, Software Requirements Specification, Schema Designing, Scripting and Testing.

3. Wrote numerous complex queries, covering all the entities and relationships, to test the database performance.

Tech… Show more

1. Worked in a team of 4 to create a SQL-based database in PostgreSQL for blood banks to manage data related to inventory, donors, patients, camps, staff, and transactions.

2. Successfully executed the entire development process from scratch including Requirements Elicitation, Software Requirements Specification, Schema Designing, Scripting and Testing.

3. Wrote numerous complex queries, covering all the entities and relationships, to test the database performance.

Tech Stack: PostgreSQL, SQL Show less

1. Analyzed neighborhood clusters of New York City (NYC), USA based on the number of coffeehouses and cafes present, the ratings, pricing and number of likes of the stores.

2. Data like store ratings, likes, price tier was collected from Foursquare Developers Places API. For neighborhood locations in NYC, 2014 New York City Neighborhood Names dataset available on New York University Spatial Repository was used.

3. Used k-means clustering algorithm to cluster 214 neighborhoods with… Show more

1. Analyzed neighborhood clusters of New York City (NYC), USA based on the number of coffeehouses and cafes present, the ratings, pricing and number of likes of the stores.

2. Data like store ratings, likes, price tier was collected from Foursquare Developers Places API. For neighborhood locations in NYC, 2014 New York City Neighborhood Names dataset available on New York University Spatial Repository was used.

3. Used k-means clustering algorithm to cluster 214 neighborhoods with prominent coffeehouses and cafes. Divided into 5 clusters, and each cluster's characteristics were discussed for the use of the target audience.

Tech Stack: Foursquare Developers API, Scikit-learn, Pandas, Python, JupyterLab Show less

1. A small-scale combination model of IMDb (Internet Movie Database) and Letterboxd for use at a personal level.

2. The front-end was created using Flutter. Users can input new entries, edit existing data, and query results based on filters.

3. Firebase Realtime Database was used in the back-end. Rules are set to support querying using multiple parameters.

Tech Stack: Flutter, Firebase, Dart, Android Studio

1. Collected MODIS data from Google Earth Engine. Kriging was performed on the dataset to account for missing values.

2. Used multiple Polynomial Regression models of varying degrees to find the relationship between (NDVI, EVI) and Carbon dioxide.

3. Performance evaluation and result analysis were done using MSE, R-squared score, and graphs.

4. Project Lead for a team of 17 members; ensured that all the deliverables were accomplished with the best possible results and… Show more

1. Collected MODIS data from Google Earth Engine. Kriging was performed on the dataset to account for missing values.

2. Used multiple Polynomial Regression models of varying degrees to find the relationship between (NDVI, EVI) and Carbon dioxide.

3. Performance evaluation and result analysis were done using MSE, R-squared score, and graphs.

4. Project Lead for a team of 17 members; ensured that all the deliverables were accomplished with the best possible results and efficiency.

Tech Stack: Scikit-learn, Pandas, NumPy, Matplotlib, Python, JupyterLab Show less

1. Designed and implemented a (2,1,2) Convolutional Code Encoder and Decoder using Viterbi Algorithm for error correction, and simulated a Binary Symmetric Channel (BSC) for performance testing.

2. Engineered (n,k) Product Code encoder and decoder utilizing Tanner Graphs, implemented error-inducing BSC and Binary Erasure Channel (BEC) algorithms, and developed LDPC code decoders for robust error correction in communication systems.

3. Conducted thorough probabilistic analysis and… Show more

1. Designed and implemented a (2,1,2) Convolutional Code Encoder and Decoder using Viterbi Algorithm for error correction, and simulated a Binary Symmetric Channel (BSC) for performance testing.

2. Engineered (n,k) Product Code encoder and decoder utilizing Tanner Graphs, implemented error-inducing BSC and Binary Erasure Channel (BEC) algorithms, and developed LDPC code decoders for robust error correction in communication systems.

3. Conducted thorough probabilistic analysis and compared theoretical and actual decoding success rates for convolutional codes and LDPC codes. Show less

1. Created detailed AutoCAD designs for a wireless land rover, applying CAD principles to ensure precise visualization and arrangement of mechanical and electronic components.

2. Designed and interpreted schematics for transmitter and receiver circuits using VLSI principles, and populated PCBs with components such as ICs, capacitors, and resistors.

3. Executed PCB fabrication processes including masking, etching, drilling, and soldering, resulting in a functional wireless… Show more

1. Created detailed AutoCAD designs for a wireless land rover, applying CAD principles to ensure precise visualization and arrangement of mechanical and electronic components.

2. Designed and interpreted schematics for transmitter and receiver circuits using VLSI principles, and populated PCBs with components such as ICs, capacitors, and resistors.

3. Executed PCB fabrication processes including masking, etching, drilling, and soldering, resulting in a functional wireless remote-controlled bot. Show less