Cloud Development Projects
The Cloud Resume Challenge
Description:
- Designed and implemented a serverless architecture utilizing AWS Lambda, API Gateway, S3, DynamoDB, and SAM Templates.
- Developed a dynamic and responsive web application using Jekyll to showcase professional experience, skills, and projects.
- Automated deployment processes with GitHub Actions for efficient and scalable infrastructure management.
- This portfolio is deployed as part of this project.
Software Projects
Clips: Video Clip Sharing Platform
Description:
- Developed a dynamic web application using Angular framework to create a video streaming platform.
- Integrated Firebase Authentication for secure user registration and login functionalities, and Firebase Storage and Firebase Realtime Database to manage video, metadata, user preferences, and video playback history, enabling efficient data storage and retrieval.
- Utilized AngularFire library to streamline communication between the Angular application and Firebase services, enhancing development efficiency and code maintainability.
Enhancing PockeyPy
Description:
- Enhanced string formatting functionalities by adding previously unavailable features mentioned in official documentation.
- Implemented Deque data structure for collections module using C++.
Extension of VetIoT
Description:
- Integrated Helion, an ML-based sequence generator, with VetIoT, a platform for testing home automation defenses.
- Created 250+ IoT applications (OpenHAB Rules) from application descriptions available for Helion and IoT defense policies for the applications for multiple IoT defense platforms.
- Built a system to translate events between Helion and VetIoT using Python, JSON, and DSLs.
Note: This is an ongoing research project, source code and other artifacts will be available upon publication.
FuzzEval: Evaluating Fuzzers
Description:
- Developed FuzzEval, an adaptable fuzzing framework, utilizing Python to evaluate fuzzers with context-sensitive inputs.
- Created customized scripts to efficiently manage fuzzing campaigns and containers.
- Evaluated the performance of 11 fuzzers across 13 different C++ libraries.
Note: This is an ongoing research project, source code and other artifacts will be available upon publication.
SeQR: Secure WiFi Authentication
Description:
- Integrated secure user authentication for connecting to enterprise wifi networks by validating server certificate through QR Codes.
- Modified LineageOS to implement the PoC of SeQR.
- Conducted thorough performance evaluations of the implementation using a customized Android NDK application and Censys dataset.
Note: This is an ongoing research project, source code and other artifacts will be available upon publication.
Customized Survey WebApplication for Amazon Mechanical Turk
Description:
- Designed and implemented a customized survey application using Flutter as the frontend and Firebase as the backend.
- Added user authentication and admin access capabiliteis to manage ongoing survey.
- Implemented real-time data synchronization to ensure immediate updates for both users and administrators.
- Empowered administrators with functionalities such as:
- Customizing and managing survey questions, including addition and deletion.
- Facilitating collaboration among administrators by allowing comments to be added alongside questions for discussion.
- Utilized Firebase Firestore for:
- Storing and organizing survey data efficiently for seamless retrieval and analysis.
- Enabling secure access controls to ensure data privacy and integrity.
- Utilized Firebase Storage for:
- Storing multimedia content such as pdf instructions associated with survey questions.
- Ensuring seamless access to other multimedia assets such as QR-code and other iamges.
Automated IoT Model Generation for NuSMV
Description:
- Developed a Python script to streamline the creation of NuSMV models from system descriptions written in a domain-specific language.
- Investigated algorithms, particularly the "Simple-path algorithm" in the NuSMV model checker, to generate intriguing system state sequences.
Note: Source code is not available for public.
PixelHarvest: Image Harvesting and Annotation Tool
Description:
- Implemented a system for collecting and providing simple annotation for Image Data.
- Developed an Android front nde using the IONIC framework, while the back end utilized Flask.
- Successfully tested on a real server, gathering 1000+ handwritten digit images.
- Designed for potential adaptation into an online classification service by hosting a trained classifier on the server side.
FileShare: File sharing over Wi-Fi.
Description:
- Implemented a system for sharing files using Wi-Fi networks in cross-os environments.
- Implemented the desktop application using JavaFX to enable cross-platform functionality.
- Implemented an android application for sharing files over Wi-Fi from android devices.
Fire Service Response Tracking
Description:
-
A fire watch system designed for:
- Calling nearby fire stations.
- Alerting nearby devices through the app.
- Tracking fire service vehicles as they respond to calls.
-
Implemented components:
- Backend database using Firebase API.
- Tracking System utilizing Google Maps and Firebase API concurrently.
- Push Notification System integrated with Firebase Cloud Messaging (FCM) API.
Note: Source code is not available for public.
Data and Machine Learning Projects
Disease Outbreaks Data Mining
Description:
- Conducted a comprehensive analysis on disease outbreaks worldwide, focusing on the correlation between socioeconomic factors and disease characteristics.
- Accumulated data on disease occurrences, socioeconomic factors, and disease natures to derive insights.
- Utilized correlation techniques including Pearson and Spearman to identify potential influences of socioeconomic factors on disease outbreaks.
- Preprocessed and analyzed data comprising 192 socioeconomic attributes and 126 disease characteristics.
- Facilitated the selection of optimal features for future Machine Learning models aimed at disease outbreak detection and prediction.
- Identified overt correlations between variables, laying the groundwork for further research into the underlying reasons behind these associations.
Single Shot MultiBox Detector in PyTorch
Description:
- Implemented a replication of Single Shot MultiBox Detector (SSD) in PyTorch.
-
Structured the project into six main parts:
- Reading, preprocessing, and loading the data.
- Model design and implementation, including loss formulation.
- Training the model with validation and hyperparameter tuning.
- Processing model outputs using Non-Maximum Suppression (NMS).
- Evaluating the model using mean Average Precision (mAP) on testing data.
- Implementing object detection in images, videos, and live feeds.
- Utilized PascalVOC 2007 and 2012 datasets for data parsing and processing.
- Built the model architecture following a referenced tutorial and structured the project accordingly.
- Trained the model with 300x300 and 500x500 images, achieving promising results.
-
Learned and implemented various techniques during the project, including:
- Atrous convolution for enlarging the field of view of filters.
- Usage of NVIDIA-Apex for half-precision training to speed up training using tensor cores.
- Developed modular code for better management and debugging during the training process.