SANJEEV KUMAR PANDEY

Experienced Delivery/Program Manager with 22 years leading large-scale transformation and application modernization projects in payments, cash management, and enterprise banking. Skilled in managing multimillion-rupee programs, cross-functional teams, vendor relationships, and ensuring regulatory compliance. Expertise includes payments, liquidity management, SWIFT, ISO20022, and core banking integration.

• Wi-Fi Based Gait Identification, Implemented an MLOps-driven deep learning pipeline using DenseNet1D and MLflow to identify human subjects with 98.3% accuracy by analyzing Wi-Fi signal perturbations.
• PM2.5 Concentration Forecasting, Developed a Spatio-Temporal Graph Neural Network (ST-GNN) and LSTM hybrid model to predict air pollutant levels, achieving an 81% reduction in test loss and 56% improvement in RMSE.
• Real-Time Object Detection Optimization, Optimized YOLO (v1–v8) architectures for high-speed inference, implementing custom bounding-box transformations and multi-scale feature extraction to enhance localized detection precision.
• Neural Architecture Engineering, Developed a modular deep learning engine from scratch, architecting generalized Convolutional, Pooling, and Dense layers in Python to support flexible kernel and stride configurations.
• High-Dimensional Data Optimization, Executed Principal Component Analysis (PCA) for image feature engineering, successfully retaining 95% of data variance while significantly reducing computational overhead and dimensionality.
• Deep Learning Gradient Analysis, Conducted rigorous mathematical modelling of Activation Functions (Sigmoid/ReLU), analysing weight/bias impact on gradient flow and 3D surface activations to stabilize network convergence.
• Enterprise MLOps Deployment, Architected end-to-end deployment pipelines using Docker and Flask to deliver containerized, real-time inference dashboards with full experiment tracking via MLflow.

• Deep Learning Architectures: Spatio-Temporal Graph Neural Networks (ST-GNN), ATGN, YOLO (v1-v8), DenseNet1D, AlexNet, LSTM, 1D-CNN, GCN (GraphSAGE).
• Neural Engineering & Optimization: Convolutional/Pooling Layer Design, Activation Function Dynamics (Sigmoid/ReLU), Multi-Head Self-Attention, Gradient Flow Analysis.
• Advanced Data Science: Principal Component Analysis (PCA) for Dimensionality Reduction, Explained Variance Analysis, Spatio-Temporal Correlation, Multivariate Time-Series Regression.
• Signal & Image Processing: WiFi CSI Phase Sanitization/Detrending, Signal Decomposition, Multi-scale Feature Extraction, Bounding-Box Transformation Algorithms.
• MLOps & Cloud Orchestration: End-to-End ML Lifecycle Management via MLflow, Containerization with Docker, Real-time Inference Dashboarding via Flask.
• Tools & Frameworks: Python (NumPy, Scikit-learn), PyTorch (torch_geometric), TensorFlow/Keras, Matplotlib, SQL.
• Evaluation & Governance Metrics: RMSE, MAE, R-Square, MAPE, Probability of Detection (POD), Critical Success Index (CSI).