Software Engineer
Grad student at Arizona State University building full-stack experiences on AWS and applying machine learning to real-world problems.
I recently completed my M.S. in Computer Science at Arizona State University, where my coursework included Data Processing at Scale, Applied Cryptography, and Semantic Web Mining.
As an AWS Developer at School Fuel, I work on React-based student and teacher dashboards for the Kairos platform, collaborating closely with backend teams to debug APIs, refine JSON contracts, and ensure reliable data flows across AWS Lambda services.
Previously, as an AI & ML Intern at Tequed Labs, I built machine learning models using Random Forests, SVMs, Linear Regression, and k-means clustering to better understand customer behavior and improve segmentation.
React-based analytical platforms for the Kairos educational ecosystem, providing real-time tracking for students and performance insights for teachers.
Overview: Building and maintaining high-performance student and teacher dashboards for the Kairos platform, a core product for School Fuel's educational initiative.
What I built: Developing responsive UI components in React, integrating with complex backend services via AWS Lambda, and ensuring seamless data flow through refined JSON contracts.
Technical depth: Focused on state management, efficient API consumption, and debugging cross-service interactions. Constant collaboration with backend teams to optimize API response times and structure.
Impact: Provides educators and students with immediate, actionable feedback on learning progress, streamlining communication and educational oversight across the platform.
Built and evaluated NLP models to distinguish computer-generated fake reviews from genuine human-written reviews across multiple e-commerce categories.
Overview: Built and evaluated NLP models to distinguish computer-generated fake reviews from genuine human-written reviews across multiple e-commerce categories.
What I built: Worked with both classical machine learning approaches such as SVM, logistic regression, and naive Bayes, and deep learning models including BERT and Fake RoBERTa.
Technical depth: The project used a balanced dataset of 20,000 fake and 20,000 real reviews, applied preprocessing and vectorization techniques, and used SMOTE-based augmentation to improve model robustness across categories.
Results: The best model achieved 98.25% accuracy, 99.42% precision, 97.10% recall, and 98.25% F1-score, outperforming the classical baselines and producing a strong benchmark for automated fake review detection.
Researched and designed a healthcare framework that combines machine learning and blockchain to improve patient services while protecting sensitive medical data.
Overview: Researched and designed a healthcare framework that combines machine learning and blockchain to improve patient services while protecting sensitive medical data.
Architecture: Studied an integrated model with an IoT data collection layer, blockchain-based transaction and access management, and a machine learning layer for anomaly detection and patient-risk analysis.
Technical depth: Explored architectures with Personal Health Care and External Record Management blockchain networks, privacy-preserving approaches such as federated learning and Hyperledger Fabric-based frameworks, and compliance considerations including HIPAA, DISHA, and COBIT.
My contribution: Served as deputy leader, overseeing team progress, quality-checking reports, organizing shared documentation, and contributing ongoing research to ensure architectural integrity.
Built a customer segmentation pipeline using unsupervised learning to help marketing teams identify shopper groups with similar spending behavior.
Overview: Built a customer segmentation pipeline to help marketing teams identify shopper groups with similar spending behavior and plan more targeted campaigns.
What I built: Preprocessed mall customer data in Python, explored distributions, visualized relationships, and applied unsupervised clustering methods to segment customers.
Technical depth: Used the elbow method to choose 5 clusters, based clustering on behavioral attributes such as annual income and spending score, and excluded gender to avoid adding weak or unnecessary separation to the process.
Outcome: Produced interpretable customer groups, including low-income/low-spending and high-income/high-spending segments, that could support marketing strategy decisions.
Role-based web platform (Student, Faculty, Admin) for automated faculty performance reporting and structured academic feedback management.
Overview: Built a web-based academic evaluation system to replace manual feedback collection and help institutions generate structured faculty performance reports.
What I built: Developed workflows for students to submit ratings, for faculty to view results, and for admins to manage classes, subjects, questionnaires, restrictions, users, and reports.
Technical depth: Designed a complex relational schema in MySQL with tables for evaluations, faculty, questions, academic terms, and class restrictions, ensuring data integrity and efficient reporting.
Why it matters: The system automated institutional evaluation at scale, drastically reducing manual overhead and organizing academic feedback loops.
Built an IoT-based real-time drowsiness detection prototype using Eye Aspect Ratio (EAR) and facial landmarks to monitor a driver’s eye behavior and trigger safety alerts.
Overview: Built an IoT-based real-time drowsiness detection prototype to monitor a driver’s eye behavior in low-light conditions and trigger safety alerts.
Architecture: Designed the system around three stages: capturing video through a night vision camera, detecting facial landmarks and eye state from frames, and correcting unsafe conditions through alerts.
Technical depth: Used computer vision techniques including face tracking, eye landmark extraction, and Eye Aspect Ratio-based feature analysis, supported by Python libraries such as OpenCV, Dlib, Imutils, and Pygame.
Outcome: The system emphasized low-latency real-time monitoring, robust low-light operation, and immediate audiovisual warnings to reduce the risk of fatigue-related incidents.
Whether you have a project, a role, or just want to connect — I'd love to hear from you. Let's build something great together.
Send Email