Salil
Hiremath

Traditional radar signal processing relied on static rule-based systems, leading to high false-positive rates and inability to adapt to evolving electromagnetic environments.

Designed a CNN-LSTM hybrid architecture for sequential signal classification, implemented real-time inference pipeline with optimized TensorRT deployment, and trained on 50K+ labeled radar signal samples.

Achieved 96.5% classification accuracy with <50ms inference latency, deployed in production defense systems processing 1000+ signals/second.

Healthcare facilities face overwhelming patient loads for routine consultations, with long wait times and overburdened medical staff for non-critical inquiries.

Built an NLP-driven symptom analysis engine using transformer models, integrated with a knowledge graph of 10K+ medical conditions, and developed an intuitive React dashboard for patient interaction.

Reduced preliminary consultation load by 35%, serving 500+ daily active users with 92% symptom-match accuracy validated by medical professionals.

Enterprise sales teams relied on manual forecasting with spreadsheets, resulting in 30-40% forecast errors and poor inventory planning decisions.

Engineered an automated ML pipeline with time-series decomposition, ARIMA/Prophet ensemble models, and XGBoost for feature-rich forecasting. Built interactive Tableau dashboards for stakeholder consumption.

Achieved R² of 0.94 on 90-day forecasts, reducing forecast error from 35% to 8%.

IoT networks face critical security vulnerabilities with 70% of devices lacking encryption, enabling man-in-the-middle attacks and data tampering in industrial settings.

Implemented a lightweight blockchain consensus protocol optimized for IoT resource constraints, paired with an LSTM-based anomaly detection system monitoring network traffic patterns in real-time.

Achieved 99.2% anomaly detection rate with 0.1% false positives, securing communications across 200+ IoT nodes with <5% computational overhead.