My research focuses on autonomous frameworks that can process high-dimensional sensor data locally and securely, reducing dependency on centralized cloud infrastructure. Key areas include:
- ⚡ Resource-Constrained Computing
- 📡 Signal Recovery in Noisy Environments
- 🔒 Privacy-Preserving Protocols
Frontiers in Digital Health [DOI: 10.3389/fdgth.2025.1685628]
- Architected a patient-centric ecosystem using Soulbound Tokens (SBTs) and Zero-Knowledge Proofs (ZKPs) for identity verification.
- Integrated a LightGBM model for emergency detection achieving 0.85 AUC.
- Validated on Polygon Mainnet with sub-4s latency and STRIDE threat modeling.
🔗 Manuscript
16th IEEE ICCCNT
- Developed a differential signal processing pipeline using dual MPU6050 accelerometers to isolate fetal kicks from maternal motion.
- Achieved a system-wide alert latency of 205ms using an ESP8266-Firebase architecture.
🔗 Manuscript
Submitted to IEEE Access (Under Review)
- Engineered a security framework for edge devices with <320 KB SRAM.
- Achieved F1-Score of 0.827 with a 190 KB memory footprint using One-Class SVM.
🔗 Manuscript
Submitted to IEEE Access (Under Review)
- Modeled spectral data from Quantum Dot Spectrometry Sensors (QDSS) to detect crop stress.
- Proposed privacy-preserving Federated Learning and Dynamic Proof-of-Stake consensus.
🔗 Manuscript
Submitted to Frontiers in AI (Under Review)
- Trained a 270M-parameter Gemma model optimized via 8-bit quantization.
- Achieved 45ms/token inference speed on Raspberry Pi 5.
🔗 Manuscript
CVRDE, DRDO (Defense R&D Org) | May 2025 – June 2025
- Architected a fault-tolerant GVA Registry Service using OpenDDS middleware for distributed Armored Fighting Vehicle systems.
- Reduced message latency by 40% and ensured 99.9% packet delivery reliability.
| Domain | Skills |
|---|---|
| Resilient Computing | Verilog, FPGA (Xilinx), ARM Cortex, ESP32/8266, Raspberry Pi 5, Quantization (Int8/NF4) |
| Intelligent Sensing | MATLAB, Python (SciPy, NumPy), FFT/Spectral Analysis, Sensor Fusion, OpenCV |
| Verifiable Security | Zero-Knowledge Proofs (ZK-SNARKs), Solidity, Threat Modeling (STRIDE), Elliptic Curve Cryptography |
| Languages & Tools | C/C++, Python, Cadence Virtuoso, LTspice, Keil uVision, LaTeX |
Solidity Polygon ZKPs
- Production-grade DApp on Polygon using Soulbound Tokens (SBT) for access control.
- Integrated Anon Aadhaar SDK for client-side ZKPs to maintain GDPR compliance.
🔗 Demo | GitHub
Flutter ESP8266 I2C Sensors
- IEEE YESIST12 Grand Finalist. Hardware-software bridge for maternal wearables.
- Implemented real-time "Store-and-Forward" logic for zero data loss.
🔗 Demo | GitHub
MATLAB Graph Theory
- Replaced standard GANs with graph-based segmentation for bandwidth-efficient transmission.
- Optimized transmission by sending only semantic maps over the channel.
Cadence Virtuoso CMOS
- High-speed CMOS design optimized for Power-Delay Product (PDP) and silicon area.
- Verified functionality via transient analysis.
- Machine Learning Specialization (DeepLearning.AI & Stanford University)
Supervised Learning, Unsupervised Learning, Reinforcement Learning
- Led development of club management platforms using React.js.
- Supervised frontend team for 5+ projects, ensuring accessibility standards.
- Mentored scouts in navigation and survival techniques; organized annual camps.
"The best way to predict the future is to create it." ✨