Prasanth Kasa

• Developed the complete software stack from scratch for a new metering device — OS bootstrapping through all application-level code for reliable data transmission
• Reduced RAM usage from 98% → 82% and cut storage by 20% by redesigning the full data propagation flow between Enphase devices
• Maintained and developed C/C++ software for Bootloader, Battery, and Micro-inverters — optimizing multi-threaded processes with Protobuf for inter-device communication

• Built predictive models to assess transaction risk within 10ms, enabling real-time fraud detection at scale
• Developed microservices in Go and Java to automate model onboarding, saving 2 weeks of man-hours per configuration
• Migrated alert and model data from distributed YAML files across multiple services to a centralized MySQL database

• Contributed to the OQLO Mobile Management solution using Firebase Cloud Messaging
• Implemented end-to-end AES-256 encryption for data at rest (AWS S3) and in transit over HTTPS/SSL
• Designed dynamic scaling with AWS Auto Scaling and Azure Load Balancer, combined with queue-based sharding for real-time peak performance

On-device AI assistant for smart glasses — real-time face recognition, scene understanding, and spatial object memory. Three parallel perception lenses running entirely on-device via Qualcomm AI Hub NPU.

Peer-to-peer communication over sound waves — no WiFi, no internet. FSK encodes text into audio tones; other devices decode via microphone + FFT. Includes mesh relay, near-ultrasonic stealth mode, and client-side Whisper AI.

CNN + TabNet pipeline predicts disaster severity from IoT/geospatial data and allocates resources in real-time by proximity and urgency. Built-in fault-tolerance with automatic fallback. Backed by published research.

Three-tier cloud RPM system: IoT wearable simulation → gateway aggregation → medical center with K-star health classification and AHP-VIKOR hospital ranking. IJIRCCE 2024 publication.