Portfolio
COSMIC-2 Satellite Data Processing System
Engineered a comprehensive automated processing pipeline for COSMIC-2 constellation thermospheric density data. The system orchestrates complex workflows including multi-day data collection with temporal overlap handling, format conversion from COSMIC CRX to RINEX, precise time synchronization across multiple satellites, and integration with professional geodetic software. Features intelligent epoch parsing, satellite constellation management, and robust error handling for operational reliability.
FPGA-Based GPS Timing Device with Microsecond Precision 🚧 Work in Progress
Designed and implemented a hardware-software co-design on an Artix-7 FPGA using Vivado and VHDL to create a precision timing device synchronized to a GPS-derived 1PPS signal. The system generates a 1 MHz microsecond pulse output (1PPuS) and serial messages containing UNIX timestamps with certainty metrics, maintaining operation during GPS signal loss through holdover mode. Developed VHDL modules for precise PPS edge detection, microsecond-resolution timekeeping, and pulse generation, while integrating MicroBlaze soft processor for NMEA data parsing and timestamp synchronization. Achieved sub-100 nanosecond PPS synchronization accuracy with ~1 μs/s holdover drift performance.
Secure Drone Mesh Network with Dynamic Routing 🚧 Work in Progress
Developing a robust drone mesh communication network where autonomous drones establish multi-hop routing paths and maintain secure communication despite mobility and link breaks. Implementing Mobile Ad Hoc Networks (MANETs) and Flying Ad Hoc Networks (FANETs) protocols using Python's asyncio for distributed, infrastructure-less communication. Features include neighbor discovery via HELLO beacons, distance-vector routing with Bellman-Ford algorithm, waypoint-based mobility simulation, and real-time topology adaptation. Future enhancements include AES/ChaCha20 encryption and integration with CARLA simulator for 3D visualization.
Ticket Classification System
Addresses a real-life challenge of Hyper Automation Support tickets, allowing the appropriate team to track down issues and resolve them efficiently. Used the K-MEAN Classification algorithm to cluster tickets into different categories.
Indoor Navigation with Augmented Reality
Combined Visual-Based Localization with Received Signal Strength Indicator for an Indoor Navigation Application. Explored Bayesian Filters for Sensor Fusion and incorporated multiuser experience with cloud-based landmark storage.
Fetal Health Classification
Machine learning model designed to reduce child and maternal mortality rates by categorizing fetal health into suspect, normal, and pathological categories with 99% accuracy using historical cardiotocograph data.