I build cool full-stack stuff, teach machines to see (hello, computer vision), and toss in some data spice when needed. When I’m not coding, I’m off hiking, traveling, or editing the moments I capture — AWS Certified DevOps Engineer and a proud Sun Devil 🔱 with a Master’s in Computer Science from ASU.
View My Work
I’m a full-stack engineer with a passion for building scalable systems that solve real-world problems. From developing Spring-based microservices to crafting dynamic frontends with React and Preact, I enjoy owning features end-to-end. My strength lies in designing backend architectures that are robust, observable, and efficient — whether it’s reducing latency in rule engines or building secure authentication flows.
I’ve worked across diverse data ecosystems, extracting business insights using BigQuery, Snowflake, and Oracle, and integrating with NoSQL stores like Cassandra and MongoDB. My approach to data is pragmatic — from ETL pipelines to real-time analytics, I care about what drives impact. I’m also experienced in observability, using tools like Splunk, Kibana, and Tableau to keep systems transparent and measurable.
At PayPal, I led initiatives that enhanced user security and saved millions by optimizing internal tools. My exposure to DevOps, containerization, and CI/CD pipelines keeps me grounded in delivery and reliability. Earlier at Nokia, I dabbled in computer vision, building lightweight models for industrial use cases.
I also published research on speaker identification using recurrence plots in Scientific Reports, pushing the boundaries of audio-based deep learning models. For me, great engineering is part code, part clarity — and I love being in the zone where both meet.
This project leverages Recurrent Plot (RP) embeddings as nonlinear features for speaker recognition, capturing unique vocal tract dynamics across air, bone, and throat microphones. It demonstrates the effectiveness of RP embeddings in unimodal, bimodal, and trimodal systems, highlighting their potential for both speaker and speech recognition applications.
Developed a custom xv6 RISC-V OS on QEMU, implementing secure boot with kernel hash checks, memory protection using PMP, and virtualization through trap-and-emulate. Integrated copy-on-write for process forking followed by lazy and on-demand memory allocation, and built complete modules for process scheduling and memory management, focusing on system isolation, performance, and security from the ground up.
Built a scalable pipeline for automobile IoT sensor data using Kafka, Cassandra, and Kafka Streams. Optimized storage with partitioning, predicate pushdown, and PostgreSQL-to-Cassandra migration. Enabled efficient query plans and real-time insights via a FastAPI backend. Developed a React-based dashboard for visualizing metrics, alerts, and sensor data severity patterns.
Designed an AI-driven Android app using tensorflow lite to make healthcare accessible on the go. Features include a virtual medical assistant, disease prediction from blood test inputs, hospital locator, smart health alerts, and emergency support. Aims to empower users with instant, intelligent medical guidance, preventive care, and real-time doctor connectivity anytime, anywhere powered by quickblox API.
Designed a location-aware retail app that streamlines customer journeys via smart notifications, QR-based store check-ins, and contactless payments. Enabled vendor onboarding and real-time personalization. Integrated BigQuery for analytics and Airflow for ETL automation. Reduced checkout friction and wait times, offering a seamless experience from entry to exit through intelligent automation and customer-vendor interface integration.
Led a student team to develop a deep learning-based computer vision application for tracking screwing patterns in manufacturing using OpenCV . Employed a CNN trained on MNIST for number recognition and created a Python QR-based alternative, achieving 100% accuracy in error detection. Deployed the application using Kubernetes for efficient scaling.
Pioneered a state-of-the-art video processing (computer vision) solution using TensorFlow's SSD MobileNet for lightweight object detection (powered by TensorFlow's object detection API). Applied scikit-image for image processing and Keras for data augmentation, followed by transfer learning (trained on COCO dataset) achieving a 94% reduction in manufacturing faults like improper screwing and board misfit.